NIST Speech Discs 16-1, 16-2, 16-3, 16-4, 16-5, 16-6, 16-7, and 16-8
A. Equipment List
B. Abbreviations used in DFW Transcripts
The Air Traffic Control (ATC) Corpus consists of nearly 70 hours of recorded communications between controllers and aircraft at three major US airports. It is really best thought of as a set of three smaller corpora, which will be referred to by the names or three-letter abbreviations of the airports where they were collected: the Dallas-Fort Worth or DFW corpus, collected at Dallas-Fort Worth International airport in Texas; the Logan or BOS corpus, collected at Logan International Airport in Boston, Massachusetts; and the Washington National or DCA corpus, collected at Washington National Airport in Arlington, Virginia.
The data were collected by Texas Instruments under contract to DARPA for the purpose of supporting research and development activities in the area of robust speech recognition for signals like ATC -- many speakers, noisy channels, relatively small vocabulary, constrained language, etc.
Each corpus consists of 20 to 25 hours of voice communications traffic between controllers and aircraft, recorded via antennas and radio receivers in the vicinity of the airport. The speech files are 8 kHz, 16-bit linear sampled data, representing continuous monitoring, without squelch or silence elimination, of a single FAA frequency for one to two hours. Each speech file is fully transcribed, and the transcripts include time markings to indicate the beginning and ending of every transmission in the sampled data.
The three corpora differ in several respects, due partly to differences in the ATC systems at the three airports, and partly to decisions made during the more than two years of design, collection, and processing. For one example, the Logan and National data sets include, for each speech file, another file indicating the amplitude of the carrier frequency signal at regular intervals; for another, the National recordings are from a different radio receiver than the other two. Thus each dataset requires some documentation of its own, and users are advised to read the description of each corpus before deciding which one best meets their needs.
In addition to the three airports, with their different local terminology and details of organization, there are several ATC functions sampled in some depth, involving takeoffs, landings, approaches, departures, and taxiing. Other functions are underrepresented or absent due to limited resources. Likewise, some important factors affecting ATC language and protocol could not be varied systematically in a corpus of this size. For example, weather conditions determine which runways are used, in which direction traffic flows, and whether Instrument Flight Rules (IFR) prevail. To sample this and other such sources of variability adequately would have required a corpus many times the size of this one.
Each airport corpus is on a separate set of CD-ROM, three each for DFW and DCA, and two for BOS. Within each corpus, the data is contained in .wav files (sampled data), .txt files (transcripts), and .cdt files (receiver carrier-detect signals, for the BOS and DCA airports only) of the form aaa_pp_n.wav, aaa_pp_n.txt, and aaa_pp_n.cdt, where aaa is the airport (dfw, bos, or dca), pp is a two-character abbreviation of the ATC position name (which may have a digit as its second character), and n is an integer index. Where there are several files from the same position, the index numbers indicate the order of collection. For example, in the DFW dataset there are three recordings, each an hour long, collected from the Arrival Radar Two position at different times, identified as dfw_a2_1, dfw_a2_2, and dfw_a2_3.
The initial specifications for the Air Traffic Control (ATC) corpus were drafted in June 1989 at a DARPA meeting attended by technical representatives of several interested government agencies and contractors.
Tom Crystal of IDA suggested sizes and dimensions for the corpus, based on his calculations of desired statistical properties. These were later circulated in a written form for review.
The fundamental consideration in determining size and dimensionality was that there be enough "trials" so that experiments could be conducted and the results accepted with a high degree of confidence. A trial, in this sense, would be an event whose occurrence could be determined by accurate recognition and interpretation of the speech: for example, the presence in the airspace of a particular aircraft (identified by its flight number), or the fact that a particular aircraft was in the process of landing, taking off, taxiing, or any other activity that is signaled by the utterances in the ATC dialogs. Thus an hour of ATC might contain 600 transmissions, consisting of 300 dialogs, involving one controller and 80 aircraft. This could provide 80 trials of flight id recognition, or 300 trials of dialog recognition, or 600 trials of utterance (or "sentence") recognition.
However, at the time of these discussions the other experimental parameters besides airports were not fixed, because not enough was known about the application in advance. For example, would an hour of data each from two different departure control positions at the same airport, or even from the same position under different weather conditions, belong in the same experiment? Or recordings of "feeder" and "arrival" at DFW, where approach control is split into two positions? If there were standard answers to these questions, they would help to determine the number of types and tokens in the corpus.
To discuss such issues, and to make final determinations of the requirements of the ATC corpus, a Steering Committee made up of representatives from ARPA (Dr. Thomas Crystal), Rome Laboratories (Dr. Edward Cupples), and BBN (Dr. Robin Rohlicek) was set up to advise TI on the collection and processing of the data.
Initial estimates were that 10 to 15 hours each would be required from as many as four civilian and one or more military airports, with emphasis on a variety of different functions and multi-function "scenarios". For example, it was decided at first to record simultaneously from multiple positions dealing with the same aircraft successively over time, so that an AI-style system could use this type of information as a constraint to improve overall performance. Thus the hypothesized detection of a particular flight communicating with the Arrival controller makes it more probable that the same flight will be detected within ten minutes communicating with Tower, and a few minutes later with Ground. Accordingly, much of the DFW data was collected in multistation one-hour sessions, in such a way that the same aircraft often appear in two or three recordings as they progress through different positions.
After the first few data collections at DFW, however, the complexity of the ATC system as a speech recognition application became more evident, and the need for homogeneous training data for speech recognition began to dominate the assessment. It was then decided that about 24 hours of data from four or five "different" functions at DFW would serve as a pilot corpus, after which the amount of data per function and the number of functions could be determined for other airports.
Based on _a priori_ estimates of traffic volume at DFW, it was felt that even the least frequent event (flight id) would occur often enough to produce reliable statistics for evaluation from a few hours of test data. At a hoped-for 50 events per hour, about 24 hours per scenario (airport), divided among about four functions, was deemed sufficient data for reliable development, testing, and evaluation in the first round.
Although the committee never reached a definitive position on what the "standard" subdivision of the ATC data should be, it did approve the collection of more data per position from fewer positions at the next two airports, in recognition of the number of other variables which can affect performance of any speech recognition-based technology on ATC. Thus the 24 hours of published DFW data come from eleven different ATC positions, which may arguably be reduced to five or four functions; the 20 hours of BOS data cover just four positions representing four functions, three of which have at least rough equivalents at DFW, and the 25 hours of DCA data cover five positions with just three functions: final approach, tower, and departure.
>From the point of view of pragmatics, therefore, DFW represents the broadest sampling and DCA the greatest depth, with BOS in between. See the individual corpus descriptions for more details.
This section describes the procedures followed in collecting all the data, as well as some of the differences from airport to airport. Refer to the sections on the individual datasets for more detail on these differences.
EQUIPMENT. At the recommendation of Rome Laboratories, TI purchased four multipurpose Radio Shack Pro-2005 scanner/receivers (RS2005), a discone antenna, a splitter-amplifier and other necessary hardware for the recording. The four receivers were needed to fulfill the requirement for parallel data at DFW from four ATC positions at once. Afterwards two of these receivers were used at Logan, and a different radio at DCA.
Before any data was collected, TI Central Research Laboratories' Tests and Measurements Lab used a signal generator and spectrum analyzer to measure the frequency response of each individual receiver, both to check their integrity and to rule out any significant differences among the four. They were all found to be within their published specifications, and very close to each other in both frequency response and harmonic distortion over the 200-4000 Hz range. Figures 1-4 show their audio levels and distortions as a function of frequency. (Figures are located in the image directory.)
For the first few sessions (about one half of the DFW data), the recording was done on video cassette recorders with digitizers at 44.1 kHz, then converted and downsampled to 8 kHz 16-bit computer files. When Digital Audio Tape (DAT) players became available, the remaining data from DFW, and everything from BOS and DCA, was recorded on DATs.
PROCESSING. All speech data are in Unix binary files, with 16-bit values at 8 kHz sampling rate, in NIST's SPHERE format. The digital files for the DFW data were produced at TI; there was a slight difference in how the data from the first three and the last two sessions were processed, which is described in section 6. The digital files for BOS and DCA were produced at BBN, courtesy of Robin Rohlicek and Manhung Siu. Their software produced the 8 kHz versions of the data which are included here, by downsampling and filtering directly from the DAT tapes.
The downsampling was done using a lowpass 256 coefficient FIR filter with the passband limited to telephone bandwidth at the upper edge. Figure 5 shows the frequency response of the filter. (Figures are located in the image directory.)
LOCATIONS. The recording sites are described separately for each airport. What they have in common is that they were in buildings with line of sight (or nearly so) from the receiving antenna to most of the transmitters (i.e., the aircraft and the ATC tower antennas). This siting was always a compromise among the many factors affecting reception. For example, in a hotel room facing the airport antennas, the arriving and departing aircraft will be on the other side of the building for at least some of the time. Close to the ATC antennas, the reception of controllers, aircraft on the runways, and aircraft on final approach will be better, but that of aircraft just beginning or at the far end of their approach turns will be worse.
The DFW corpus was collected over 5 sessions, each at a different place and time, while the Logan and DCA corpora were collected from one location over a period of consecutive days.
PROCEDURE. The DFW collection was designated as a "pilot corpus" from the beginning, and minor changes in collection protocol and even equipment were allowed between recording sessions for this reason. They are documented in the section on the DFW corpus. Offsetting this is the fact that there was far more DFW data collected than could be published, and the "worst-sounding" recordings (those with the most out-of-band interference, usually) were set aside. On the whole the DFW data is probably no less uniform or more difficult to deal with than the other two sets.
The BOS recordings used the same receivers, antenna, and DAT recorders as the later DFW data. They were made from a similar location, but in three days of nearly continuous recording rather than in sessions spread out over time. The Logan airport layout and its ATC functions are very different from DFW, which implies changes in vocabulary, phrasing, and pragmatics. This is reflected in the choice of only four functions, none of which are exactly like any of the DFW functions. Files are also longer, typically two hours rather than one. The processing of the data also changed somewhat: BBN provided software to downsample and filter the 48 kHz DAT signal to produce the 8 kHz speech files, avoiding the resampling step required with the DFW data.
The DCA data is a significantly different signal from DFW and BOS; a new, aircraft-quality receiver and matched antenna were used to improve selectivity without loss of sensitivity. The location, on the roof of the ARPA building a few miles north of National Airport and close to the flightpath in a south flow, was comparable to the Hilton location for DFW. Its only serious disadvantage was that signals from aircraft on the ground were sometimes temporarily blocked when they taxied behind the terminal buildings. As in the case of Logan, DAT recordings were made more or less continuously over a three day period without moving or changing equipment. The ATC functions were more similar to DFW than to BOS, but not identical to either.
ATC FUNCTIONS AND POSITIONS. The role of an air traffic controller in a major metropolitan airport terminal is defined in terms of the entire system for that airport, a subject too far afield for this document. Nevertheless a brief discussion of the terms may be useful. The major terminal-based functions are as follows:
Ground Control (GC): Typically guides aircraft from company gate or terminal area via taxiways to last ramp before takeoff, and from the first off-ramp to the terminal area after landing.
Local Control (LC), also called Tower: Typically handles aircraft on final approach, clears them to land, hands them off to Ground Control; also takes aircraft from Ground Control, clears for takeoff, then hands off to Departure Control.
Departure Radar Control (DR), also called Initial Departure (ID) at BOS, Departure Low at DCA: Typically handles departing aircraft from shortly after takeoff to a certain altitude, where they are handed off to either a high altitude Departure Control (at DCA) or to a "Transition" controller (at BOS), or immediately to an enroute controller (at DFW).
Arrival Radar Control (AR), also called Final: Typically handles aircraft during final approach maneuvers within the greater terminal radar control area; hands off to Local Control within a few miles of runway for landing. May be split into Feeder and Final (normally at DFW, rarely at DCA).
Feeder Control (FE, FW for East and West at DFW): Typically guides aircraft approaching from various directions into a line for the downwind leg of final approach, then hands off to Final (AR). Same general type of communications as Final, but first contact may be 30 miles out from runway.
Transition Control (SM) also called Plymouth at BOS: Typically guides both departing and arriving aircraft at some distance from the airport, like Departure High for departing aircraft at DCA, and like Feeder for arrivals at DFW.
Positions vs. Functions: These functions may each be performed by more than one controller at an airport, even at the same time. At DFW, where parallel runways permit most operations to proceed independently on two sides, there are one, two, or three controllers handling departures, depending on a variety of factors; so at a particular time, Departure Radar Three (DR3) may be the only one on the air. Let us use the term "position" for the local name associated with a particular frequency manned by one controller. This name generally signifies the primary responsibility of the controller on that frequency: Ground Control East at DFW, or Plymouth Transition at Logan (handling both incoming and outgoing traffic near the intersection called Plymouth.)
Note also that, although positions may be named for their principal functions, each controller actually "owns" a defined three-dimensional airspace, and the "arrival" and "departure" positions also deal with any aircraft passing through their airspace, even though it may not be arriving or departing their airport. For example, the Feeder East position at DFW often handles flights on their way to and from Love Field in Dallas, in addition to those arriving at DFW.
Combined positions: It is not unusual for a controller at one position to take over what would be done by another controller at another position. In fact, this is routine in smaller airports and at off-peak hours for reasons of efficiency. Particularly during slow periods, controllers may "combine" positions without notice, so that they can be heard on two frequencies. A receiver set to one of these frequencies will hear all the controller's transmissions but only some pilots' answers, since the others are communicating on another frequency. This commonly happens, for example, with the departure positions at DFW.
The ATC positions represented in these corpora are standard in their respective airports, and were recorded deliberately during relatively busy traffic periods. This strategy was dictated by the need for training data. Experiments with enroute and military ATC positions, for example, yielded less than one quarter as many transmissions as these busy metropolitan terminal positions, so that enormous numbers of hours would have to be collected and condensed in order to provide training data for most algorithms.
However, even major airports are organized in different ways, so one must not assume that similar sounding positions, even those with the same name, perform exactly the same function in different airports. For example, at DFW there are two tower controllers, one handling traffic landing and departing on the east side runway(s), one handling the same on the west side. In Boston, where the major runways intersect, there may be two tower controllers on the air, one of whom handles mostly takeoffs and the other mostly landings. See the descriptions for each airport below.
Air traffic control speech is difficult for most laymen to understand for several reasons: channel limitations, noise, rapid pace, and specialized jargon, for example. Most important, however, is the absence of contextual constraints; there actually are numerous constraints, but only the participants know them and can exploit them. For example, knowing the usual downwind path of arriving aircraft and the current position of a particular one makes the controller's direction (e.g., "turn right heading one zero seven") natural and obvious; without that information almost any numbers could occur. Although lay transcriptionists can and do learn to understand and transcribe it, the process is long and difficult.
The transcriptions of the ATC recordings were therefore done by two former air traffic controllers from the Dallas/Fort Worth area who had spent many years working in the DFW ATC system. They were experienced in transcribing ATC traffic for legal and governmental purposes, such as investigation of accidents and incidents, planning and revision of local ATC systems, etc. Such transcripts are designed for human reading, and normally follow Federal Aviation Administration (FAA) practices, which constitute a _de facto_ standard in this area.
In addition to errors which would not necessarily need correction in a document used for legal or investigative purposes, such as spelling mistakes, the original transcriptions had a number of inconsistencies and arbitrary usages: hesitation sounds (spelled ah, uh, eh, ahh), attempts to represent pronunciation (climbin', awright, 'kay), and a variety of terms to denote problems ("unintelligible," "sounds like," "blocked transmission," etc.).
The FAA format was therefore unsatisfactory for electronic processing of the text, and a set of conventions was established in cooperation with BBN, which was the first intended consumer of the data. These conventions were developed over a period of several months during the pilot phase of collection. Some of them were then imposed on the transcribers, while others were produced algorithmically from the transcripts using awk scripts or macros during the process of review and quality control at TI.
BBN requested that each transcript be in the form of a Lisp list with four kinds of objects: a header, a tail, stand-alone comments, and transmissions. Transmissions are also grouped into dialogs bounded by extra newlines for human readability, but this is not part of the specification and not guaranteed to be consistent.
The specification calls for every transmission to contain:
The TEXT contains transcribed words in upper case with no punctuation other than possessive apostrophes, and with certain reserved expressions permitted in parentheses, such as SHORT PAUSE, LONG PAUSE, UNINTELLIGIBLE.
The following passage contains examples of all these kinds of objects, including both transmission-internal and stand-alone comments:
((TAPE-HEADER "DFW ATC FEEDER EAST, 119.5 MHZ, 1-7-91, 1550 CST")) ((FROM AAL586) (TO FE-1 ) (TEXT AND AMERICAN FIVE EIGHTY SIX IS WITH YOU AT AH ELEVEN) (TIMES 4.69 7.80)) ((FROM FE-1) (TO AAL586 ) (TEXT CALLING APPROACH SAY AGAIN) ((FROM FE-1) (TO TWA319) (TEXT T W A THREE NINETEEN DEPARTURE AT TWELVE O'CLOCK SEVEN MILES EASTBOUND CLIMBING TO ONE ZERO THOUSAND) (TIMES 21.42 25.32)) ((FROM FE-1) (TO DAL265) (TEXT DELTA TWO SIXTY FIVE MAINTAIN SIX THOUSAND CONTACT APPROACH ON ONE ONE NINER POINT FOUR) (TIMES 551.69 555.60)) ((FROM DAL265) (TO FE-1) (TEXT ONE NINETEEN FOUR DELTA) (TIMES 555.81 557.96) (COMMENT "SOUNDS LIKE SIXTEEN")) ((FROM FE-1) (TO AAL633) (TEXT AMERICAN SIX THIRTY THREE HEAVY DESCEND AND MAINTAIN FIVE THOUSAND APPROACH NOW ONE ONE NINER POINT FOUR (LONG PAUSE) AMERICAN SIX THIRTY THREE HEAVY CONTACT APPROACH ONE ONE NINER POINT FOUR) (TIMES 2429.33 2433.21)) ((FROM AAL633) (TO FE-1) (TEXT NINETEEN FOUR AMERICAN SIX THIRTY THREE HEAVY) (TIMES 2435.83 2438.55)) ((COMMENT "SOUNDS LIKE BACKGROUND CONVERSATION--COMING DOWN-- AMERICAN NINE OH")) ((FROM FE-1) (TO AAL383) (TEXT AMERICAN THREE EIGHTY THREE CONTACT APPROACH ONE ONE NINER POINT FOUR) (TIMES 2526.45 2528.90)) ((FROM UNK) (TO UNK) (TEXT (UNINTELLIGIBLE)) (TIMES 2529.39 2530.23) (COMMENT "SOUND OF INTERFERENCE")) ((FROM FE-1) (TO N51HC) (TEXT CITATION FIVE ZERO ONE HOTEL CHARLIE REGIONAL APPROACH ROGER VECTORS LOVE) (TIMES 2530.42 2533.32) (COMMENT "CONTROLLER APPEARS TO MISSTATE CALL SIGN -JJG")) ((FROM N51HC) (TO FE-1) (TEXT RIGHT ON) (TIMES 2534.02 2534.48)) ((TAPE-TAIL "END OF TAPE"))There were also conventions established for flight identification numbers and a number of other ATC-related phenomena such as the names of intersections. Some of the technical terms for DFW are given by way of illustration in Attachment B, "Abbreviations Used in Transcriptions," prepared by one of the ATC consultants. Similar material for other airports can be obtained from the FAA.
The DFW data is contained in 24 .wav files, each approximately one hour long, and 24 corresponding .txt files of transcripts. The data was collected over a period of about a year (March 1990 to February 1991) from several locations on or near the airport grounds.
EQUIPMENT. The DFW recordings in the first three sessions were made using a discone antenna, a four-way splitter, up to four scanner receivers, two stereo digitizers, and two video cassette recorders (VCRs). For the last two sessions, Sony TC75ES Digital Audio Tape Recorders (DATs) were substituted for the digitizers and VCRs. See the description in Attachment A.
PROCESSING. The data in the first three sessions was collected digitally on VCR tapes. It was later converted to computer files by playing back the analog version, resampling with a DSC converter and low-pass filtering with a Vax 780. The DAT recordings from the other two sessions were processed by filtering and downsampling with a PC-based DAT interface in the Consumer Products speech processing laboratory. Since the sampling rates of the DATs (48 kHz) and the VCR digitizer (44.1 kHz) are both so high, any differences in the DFW dataset due to processing should be minimal.
LOCATIONS. A diagram of the airport appears in Figure 6. (Figures are located in the image directory.)
The DFW data was collected in six sessions, from guest rooms in three different hotels in the vicinity of the airport. The "Tape Header" item at the beginning of each transcript gives the date and time of recording. The recording session dates and locations were:
1 -- 03/01/90 Hyatt E (E side), antenna on roof 2 -- 03/28/90 Hilton (SW side), antenna inside 3 -- 05/18/90 Hyatt W (W side), antenna inside 4 -- 05/19/90 Hyatt W, antenna inside 5 -- 01/07/91 Hyatt E, antenna inside 6 -- 02/13/91 Hyatt E, antenna insideThe Hyatt East and West are twin buildings actually both on the airport grounds, less than a quarter mile from the control tower, and fronting the two principal runway complexes on either side of the airport. The Hilton is approximately two miles north of the tower, directly under approaching aircraft in a south flow. In the following schematic layout, the runways are labeled for a north flow; in a south flow, runways 36 Left and 36 Right (facing 360 degrees, due north) become 18 Right and 18 Left respectively (180 degrees, due south), while 35R becomes 17L, 35L becomes 17R; 31L and 31R become 13R and 13L.
N o ^ r | t | h Hi || || \ || || \ \ || || \ Hi = Hilton \ || HW HW || \ HE, HW = Hyatt E & W \ || o || \ | = Runways \ || || \ o = Control Tower \ || || \ 31L\ || || 31R || || || || 36L||36R 35L||35RPROCEDURE. During the first recording session, the concern was to get a strong enough signal from aircraft which could be as far as 25 miles away. The antenna was placed on the hotel rooftop, and an amplifier was inserted before the splitter to compensate for losses due to the long antenna cable and four-way division of the signal. However, a review of the data showed that episodes of interference from other signals, probably not ATC, were the principal problem, and that sensitivity was more than adequate -- airplanes on Feeder East as far as 25 or 30 miles away seemed quite audible.
Engineers from the TI Antenna Group were then asked to characterize the entire receiving setup, including the antenna, splitter, connectors, and cables. Their report, which runs to about 100 pages including charts and tables, concludes that problems with out-of-band signals are inevitable because of the broad frequency range of the multiband scanner-receivers. As many as 25 harmonics within the ATC band could be detected from a digital pager signal broadcast from the Hyatt hotel roof, for example. On the other hand, they conclude that sensitivity is sufficient for aircraft signals up to 20 miles away, even through buildings, and that therefore an indoor placement of the antenna, e.g., at a window with line of sight to most of the traffic, might make the interference problem less severe by attenuating undesired signals, especially distortion products from out-of-band transmissions.
This procedure was therefore followed for all the remaining sessions: recording equipment was set up in a room on or near the top floor, with one or more windows facing the general direction of the majority of the expected traffic. The antenna was set up in the room near the window and connected to the splitter directly without an amplifier.
Of the eight hours collected in the first session (antenna outside, amplifier in line), the four hours from the west side, which contained many more incidents of interference, were discarded, and the four from the east side, which had fewer, were retained.
The ATC functions recorded at DFW were Feeder, Arrival, Local, Ground, and Departure. DFW operates in some respects almost as two separate airports, symmetrical about a north-south dividing line, with the principal traffic flow being either to the south or to the north, depending on prevailing weather patterns. Thus "Feeder East" and "Feeder West", "Arrival Radar One (East)" and "Arrival Radar Two (West)", "Local Control East" and "Local Control West", "Ground Control East" and "Ground Control West" are nearly mirror image functions when both sides of the airport are active.
This symmetry is not complete; in addition to the principal north-south runways, which are called 17 and 18 in a south flow, 35 and 36 in a north flow, there are two others flanking these, called 13 and 31, which are used during very busy periods for either arrivals or departures. Control of this traffic falls mainly to one or the other of the sides. Nevertheless, it seemed advisable to collect in roughly equal proportion from both sides of these "symmetrical" functions in order to get data representative of the airport as a whole. The vocabulary associated with each side, for example, will contain different navigational fixes, runways, and even airline names.
Departure radar (DR) control is still less symmetrical. There are three DR positions which may be active in any combination, with DR1 tending to handle aircraft departing on the east side (which are often also eastbound) and DR3 those on the west, and DR2 (on the few occasions when all three are used) handling traffic from either runway which is neither eastbound nor westbound. But Departure positions are often combined so as to occupy only one controller, or a controller is added for brief peak periods, so it is difficult to collect long stretches with a single controller and dense traffic. All three departure positions are represented in the corpus.
TABLE OF CONTENTS.
The following table shows the DFW data arranged according to major external factors--locations, recording sessions (with equipment and location fixed), and ATC positions. On the right, the ATC positions are grouped at two levels: the nearly-identical East and West pairs and the three Departures (explained above) at one level, then the Feeder and Arrival functions, which are both "arrival" positions in a broader sense, but are split at DFW and a few other airports. All recordings are approximately one hour in duration.
SESSIONS / 1 2 3 4 5 5 5 6 / hours/function ============================================================= FE | XX | | | | XX | | | | ----|----|----|----|----|----|----|----|----|> 5 Feeder \ FW | | XX | | XX | | | XX | | \ ----|----|----|----|----|----|----|----|----|------- > 10 Arrival AR1 | XX | | | | XX | | | | / A ----|----|----|----|----|----|----|----|----|> 5 Arrival/ AR2 | | XX | | | | | XX | XX | T ----|----|----|----|----|----|----|----|----|---------- LCE | XX | | | | | XX | | | C ----|----|----|----|----|----|----|----|----|> 5 Local LCW | | XX | XX | XX | | | | | ----|----|----|----|----|----|----|----|----|---------- S GCE | XX | | | | | | | | T ----|----|----|----|----|----|----|----|----|> 3 Ground A GCW | | XX | XX | | | | | | T ----|----|----|----|----|----|----|----|----|---------- I DR1 | | | | | XX | XX | | XX | O ----|----|----|----|----|----|----|----|----|\ N DR2 | | | | | | | XX | | >6 Departure S ----|----|----|----|----|----|----|----|----|/ DR3 | | | XX | | XX | | | | ----|----|----|----|----|----|----|----|----|---------- Date 3/1 3/28 5/18 5/19 1/7 1/7 1/7 2/13 Start 1400 1345 1815 0930 1552 1705 1800 1030 - End -1500 1445 1915 1030 1652 1755 1900 1220 Times CST CST CDT CDT CST CST CST CST SESSION DATES & LOCATIONS: 1 -- 03/01/90 Hyatt E, antenna on roof 2 -- 03/28/90 Hilton SW, antenna inside 3 -- 05/18/90 Hyatt W, antenna inside 4 -- 05/19/90 Hyatt W, antenna inside 5 -- 01/07/91 Hyatt E, antenna inside 6 -- 02/13/91 Hyatt E, antenna inside ABBREVIATIONS: FE, FW: Feeder East and West. AR1, AR2: Approach control for principal E and W runways. LCE, LCW: Local (Tower) Control for principal E and W runways. GCE, GCW: Ground Control for E and W sides, taxiways, terminals. DR1, DR2, DR3: Departure Control for traffic heading E, N/S, W.
The Logan data is contained in 11 .wav files, 11 corresponding .txt files of transcripts, as well as another set of 11 .cdt files which are described below. As mentioned in Section 3 above, it differs from the pilot data collected at DFW in a number of ways, some of them significant and others not.
There were a number of changes in the collection methodology. DFW was a pilot project, and certain dimensions of ATC were explored there which were not worth pursuing further. For example, collection from multiple ATC functions for the same flight during its landing or takeoff, or recording of Ground Control or Enroute Sectors, were judged less valuable than more hours of a few basic functions, such as Tower, Arrival, and Departure.
Collection in one hour sessions was also criticized by the users of the DFW data, for the following reason. Whenever a recording session ends, any flights still being worked on that frequency are, as it were, "cut off" before they have communicated with the controller the usual number of times; the fewer examples of a flight there are, the more difficult it is to identify that flight by speech or voice recognition. There is a handful of these cut off flights, i.e., flights with fewer than normal transmissions, at the very beginning of a recording session and a like number at the very end. Since the number is approximately a constant for a given frequency, at least during busy times of day, longer recording sessions were requested so as to reduce the relative effect of this truncated traffic. The Logan recordings were therefore generally longer, up to a limit of two hours, the length of a DAT tape. However, there were still sessions where this length could not be justified: traffic might drop off on one of the channels being collected on a DAT, frequencies might be combined, etc., making it more sensible to stop and change stations after one hour, or to trim off the leading or trailing parts of a recording.
All the Logan data was collected in a period of two days, June 26 and 27, 1991. A total of about 27 hours were recorded, of which nearly 20 were acceptable for final use. Reasons for exclusion were: not enough traffic, unacceptable noise or distortion on the recording, ATC function not sufficiently well represented overall.
EQUIPMENT. The BOS recordings were made using the same discone antenna, and two of the four scanner receivers used at DFW. These had been specially modified to provide a second output signal, in addition to the audio. The reason for the modification was a change in the specifications of some contractual work being performed by BBN using the ATC data. This change required that BBN use the radio carrier signal as well as the audio. Circuitry was added to the receivers at Rome Laboratories to detect the amplitude of the carrier signal being received and to use it to control a voltage controlled oscillator. The oscillators were chosen so that their lowest frequency, emitted when no carrier was detected, was between 8 and 10 kHz, and their highest frequency, emitted when the carrier signal was at maximum amplitude, was between 20 and 22 kHz. Table 1 indicates the power and frequency relationships for two of the receivers. For a given ATC frequency, both the audio signal and the carrier-detect FM tone were recorded on one track of a Sony TC75ES Digital Audio Tape Recorder (DAT). Thus two frequencies could be recorded at the same time, using two radios and one DAT.
Power-Frequency Relationship of Carrier Detect Signal for Two Receivers:
Attenuation: 0 dB no modulation Frequency: 118.5 MHz Type - AM SN 827062 SN 803676 --------- --------- Antenna Power Frequency Antenna Power Frequency 1 uV 8.970 KHz 1 uV 9.657 KHz 3 uV 9.500 KHz 3 uV 10.939 KHz .01 mV 10.141 KHz .01 mV 12.370 KHz .03 mV 10.712 KHz .03 mV 13.262 KHz 0.1 mV 11.180 KHz 0.3 mV 13.312 KHz 0.3 mV 11.543 KHz 1 mV 13.373 KHz 3 mV 11.593 KHz 3 mV 13.748 KHz 10 mV 11.630 KHz 10 mV 14.491 KHz TABLE 1PROCESSING. By common agreement the DAT recordings were copied and sent for processing to BBN, since they were to use the new signal format in their research. The processing involved detecting the FM tone and recording its frequency into a 100 Hz, 32 bit computer file, as well as filtering and downsampling the audio to an 8 kHz, 16-bit computer file.
LOCATION. A diagram of the airport is given in Figure 7. (Figures are located in the image directory.)
The BOS data was collected from a fifth floor room in the Hilton Hotel just outside the airport terminal. The window and balcony of the room overlooked the tower and runway complex, which were less than half a mile distant. The antenna was placed on a small balcony outdoors, with about 2 meters of coax cable leading to a "Y" connector and thence to the two radios. The recording conditions were thus quite similar to those at DFW.
PROCEDURE. The ATC functions at Logan are as follows:
Final Approach (F1), comparable to the AR1 and AR2 positions at DFW, controls final approach to a few miles from the runway.
Initial Departure (ID), comparable to the DR1, DR2 and DR3 positions at DFW, controls departures from a few miles off the runway through an altitude of about ten thousand feet.
Local East (LE) and Local West (LW) are the tower positions; however, they are significantly different from DFW in that they may, for long periods, handle only landings or only takeoffs. The amount of traffic, the allocation of traffic between the shorter and the longer runways, and between the parallel and intersecting runways, and other factors affect this division of labor between LE and LW. For the corpus we chose two recordings of LW, one two hours long and one an hour long. During both periods, LW was mainly controlling takeoffs on Runway 22 Right; however, during part of the second session Runway 22 Left is closed for inspection and LW controls landings as well as takeoffs.
Lincoln (SL), Plymouth (SM), and Rockport (SR) are called "Transition" positions. They combine the functions of Feeder and Departure at DFW, in the sense that they control both arriving and departing traffic in the area. They handle arrivals before they reach F1, and departures after they leave ID. Which one(s) of the three Transition controllers are on the air depends in part on the traffic pattern in effect. On the two days of recording, with a predominately southwest flow, the Plymouth Transition (SM) was most active. The other two positions were either not on the air or handled much less traffic.
TABLE OF CONTENTS.
This table shows the tape identifier, ATC function, number of minutes, and the approximate number of distinct flights and transmissions, for each of the 11 Logan files.
Tape ID | ATC Pos | Mins |#Flights |#Transmissions --------|---------|------|---------|-------------- log_f1_1| F1 | 120 | 75 | 1028 --------|---------|------|---------|-------------- log_f1_2| F1 | 120 | 73 | 1084 --------|---------|------|---------|-------------- log_id_1| ID | 120 | 88 | 704 --------|---------|------|---------|-------------- log_lw_1| LW | 120 | 173 | 851 --------|---------|------|---------|-------------- log_sm_1| SM | 120 | 69 | 693 --------|---------|------|---------|-------------- log_id_2| ID | 100 | 81 | 602 --------|---------|------|---------|-------------- log_f1_3| F1 | 120 | 82 | 948 --------|---------|------|---------|-------------- log_sm_2| SM | 120 | 65 | 697 --------|---------|------|---------|-------------- log_lw_2| LW | 60 | 80 | 434 --------|---------|------|---------|-------------- log_id_3| ID | 64 | 43 | 266 --------|---------|------|---------|-------------- log_f1_4| F1 | 101 | 71 | 969 --------|---------|------|---------|-------------- KEY TO POSITIONS: F1: Final Approach ID: Initial Departure LW: Local West SM: Plymouth Transition
The DCA data is contained in 17 .wav files, 17 corresponding .txt files of transcripts, and 17 .cdt files with carrier frequency information. The data was collected over a period of three days, May 26-28,1992, from a single location in northern Virginia.
EQUIPMENT. The DCA recordings were made with a different radio receiver and antenna than the other two datasets, although a version from one of the same RS2005 receivers was also recorded.
Although it was never the purpose of this project to obtain "optimal" voice recordings, experience with the data from DFW and Logan, advice from the TI Antenna Group engineers, and comparisons with FAA recordings, led us to explore alternatives to the broadband scanner-receivers originally recommended. Our suspicion was that the extremely wide receptive band featured by these radios, though it provides easy access to many communications functions at a reasonable price, might be allowing unwanted signals to mix at some stage with the ATC audio. Some pilot recordings (not included in this corpus) were made at DFW with a single antenna feeding both a RS2005 and a receiver designed to admit only frequencies in the ATC range (approximately 118 to 132 MHz). These recordings, made in parallel with the output of the RS2005 scanning receivers, proved that a very common source of interfering noise in the recordings from the RS2005s was the reception of signals from outside the ATC band, including harmonics of such carriers as digital pagers and FM radio broadcasts. This led to the choice of an ATC receiver for the last corpus.
The radio chosen was a Bendix-King Model KY-97A, commonly used in small aircraft. An antenna designed for use in ATC communications was also used, a portable omnidirectional VHF Model AV-1 from Antenna Specialists. The antenna output was split and sent to both a RS scanner receiver and the BK Model KY-97A.
Monitoring of the audio during the recording sessions confirmed that the BK audio suffered significantly less interference from unwanted signals than the RS. In fact, due to an apparently powerful and leaky broadcasting antenna nearby, there was one ATC frequency (119.85) which could almost never be understood through the RS receiver, but remained free of interference through the BK.
The receiver outputs were passed through the same mixer as at Logan, except that the second receiver (the BK) did not have a carrier detection signal mixed with its audio. Thus the RS2005 carrier signal and audio were mixed and recorded together on one track of the DAT (the same Sony TC75ES used in the other two airports), and the BK audio alone was recorded on the other. In this way, comparisons could be made between the two receiver audio signals, while either one could be combined with the carrier signal without loss of time registration (see below). For the published LDC corpus, however, the signals chosen were the BK recording and the carrier detection signal, reduced in a manner similar to the Logan data, as described below.
PROCESSING. As with the Logan data, the DAT recordings were copied and sent to BBN for extraction of 8 kHz .wav computer files, to insure that the filtering and format would be identical.
Recovery of the carrier-detect signal was done at NIST, using the identical software as that used at BBN and, as at BBN, using a Silicon Graphics Indigo workstation to read the DATs.
For the type of DATs used, there is not a unique zero point. Therefore, there was a time offset between the BK audio and the carrier signal recovered at NIST. The Sony audio was also recovered at NIST. By comparing it with the BK audio, the time offset was determined. Samples were either deleted or added at the beginning of the carrier detect files so that they were in time alignment with the BK signal. The added samples are all zeroes. Zeroes were also added at the end, if necessary, to make the time durations of the carrier files approximately equal to those of the BK files. Users should be aware of this possible presence of zero values. Where speech occurs at the very beginning of a file, an initial segment included in the BK file may have zero values in the carrier file.
LOCATION. A diagram of the airport is given in Figure 8. (Figures are located in the image directory.)
The DCA data was collected over a three day period from the computer room on the tenth floor of the ten-story ARPA headquarters building at 3701 Wilson Blvd. in Arlington. The antenna was placed on the rooftop, and since less than 30 feet of BNC cable was required to reach the receivers in the computer room, no amplification was used.
The ARPA building is almost directly north and slightly west of the DCA airport, approximately 3 miles from the end of the main runway. The runways in use during a north flow, which prevailed for the three days of recording, are 3, 33, and 36, so all departing flights pass within a few miles of the building on takeoff. Arriving flights approach from the other side of the airport, but many of them first make the downwind leg of their approach directly to the west of this point.
PROCEDURE. Only three ATC functions were recorded at National Airport, though two of them had East and West positions:
Final One (F1) and Final Two (F2), comparable to the AR1 and AR2 positions at DFW, control final approach to a few miles from the runway, depending on whether the traffic was approaching from the West (F1) or the East (F2).
Departure Low West (DR1) and Departure Low East (DR2), comparable to the DR1, DR2 and DR3 positions at DFW, control departures from a few miles off the runway through an altitude of about ten thousand feet.
Local (LCL) is the tower position, handling both takeoffs and landings on two or three runways at once.
As with the Logan airport data, an attempt was made to record up to two hours at a time; shorter sessions are due to lack of traffic, or the need to switch to another frequency for a known period of busy traffic. Jim Owens, consultant to the project, had determined the "rush" and "push" times for DCA traffic in advance from FAA personnel at the airport.
The following table contains the same information as the Logan table: tape identifier, ATC station, length of recording, approximate number of distinct aircraft, number of distinct transmissions.
TABLE OF CONTENTS
ATC at Washington National Airport (DCA)
Tape ID | ATC Pos | Mins |# Flts|# Transmissions --------|---------|------|------|--------------- dca_f1_1| F1 | 56 | 26 | 443 --------|---------|------|------|--------------- dca_d1_1| DR1 | 90 | 38 | 396 --------|---------|------|------|--------------- dca_lc_1| LCL | 100 | 137 | 1093 --------|---------|------|------|--------------- dca_f1_2| F1 | 120 | 29 | 485 --------|---------|------|------|--------------- dca_lc_2| LCL | 120 | 142 | 1167 --------|---------|------|------|--------------- dca_f2_1| F2 | 60 | 24 | 383 --------|---------|------|------|--------------- dca_lc_3| LCL | 60 | 48 | 358 --------|---------|------|------|--------------- dca_d1_2| DR1 | 70 | 33 | 313 --------|---------|------|------|--------------- dca_d2_1| DR2 | 60 | 33 | 274 --------|---------|------|------|--------------- dca_lc_4| LCL | 56 | 81 | 535 --------|---------|------|------|--------------- dca_f2_2| F2 | 60 | 38 | 445 --------|---------|------|------|--------------- dca_lc_5| LCL | 120 | 118 | 939 --------|---------|------|------|--------------- dca_d2_2| DR2 | 61 | 30 | 329 --------|---------|------|------|--------------- dca_d1_3| DR1 | 120 | 50 | 539 --------|---------|------|------|--------------- dca_f2_3| F2 | 120 | 52 | 610 --------|---------|------|------|--------------- dca_d1_4| DR1 | 120 | 60 | 537 --------|---------|------|------|--------------- dca_lc_6| LCL | 120 | 119 | 928 --------|-----|------|------|--------------- KEY TO POSITIONS: F1: Feeder & Final One (West) -- 119.85 MHz F2: Feeder & Final Two (East) -- 124.20 MHz DR1: Departure One Low (West) -- 118.95 MHz DR2: Departure Two Low (East) -- 126.55 MHz LCL: Local (Tower) -- 119.10 MHz
Radio Shack Model Pro 2005 Scanner/Receivers (4)
Radio Shack Discone Antenna, Cat. No. 20-013
Bendix/King Model KY197A Aircraft Radio
Antenna Specialists Model AV-1 VHF antenna
VCR-based Recording systems (2 x 2-channel):
Panasonic A100 Video Tape Recorder
Panasonic AG-6300 Video Tape Recorder
Nakamichi DMP-100 Digital Mastering Processors (2)
DAT Recording systems:
Sony DTC-75ES Digital Audio Tape Recorder (2)
Mixer (used with modified 2005 receivers):
TASCAM M-18 Mixer
The following are lists of abbreviations common in FAA terminology. They are taken from sources for the DFW ATC system. By understanding how these flight IDs, military designators, intersection names, and other terms are used at DFW, one can readily identify and understand how the corresponding terms are used at Logan and National Airports as well, even without knowing the exact referents of each.
In addition to the air carrier designators, which form part of the flight IDs used in the transcripts, also listed are navigation aids, including their names and three letter designators, intersections along airways based on the navigation aids, Standard Terminal Arrival Routes (STARS) and intersections designated on these routes and intersections on the Instrument Landing Systems (ILS) serving D F W airport.
These aids, intersections, and routings are frequently referred to in Air Traffic communications by the D F W ATCT facility.
1. AIR CARRIER 3 LETTER IDENTIFIERS
Three letter air carrier company designators are assigned to all air carriers by the F A A using ICAO standards. When used in conjunction with the flight number they serve as the aircraft identifications in Air Traffic Communications.
3-LETTER DESIGNATOR COMPANY CALL SIGN ------------------- ------- --------- AAL (+ flight number) AMERICAN AIRLINES AMERICAN COA CONTINENTAL AIRLINES CONTINENTAL DAL DELTA AIRLINES DELTA EAL EASTERN AIRLINES EASTERN PAA PAN AMERICAN WORLD AIRWAYS CLIPPER MTR METRO FLIGHT INC. METRO ASA ATLANTIC SOUTHEAST AIRLINES ASEA CPL CHAPARRAL AIRLINES CHAPARRAL BAW BRITISH AIRWAYS SPEEDBIRD UAL UNITED AIRLINES UNITED TWA TRANS WORLD AIRLINES T W A MEP MIDWEST EXPRESS MIDEX MID MIDWAY AIRLINES MIDWAY MXA MEXICANA AIRLINES MEXICANA UPS UNITED PARCEL SERVICE UPSCO FDX FEDERAL EXPRESS CORPORATION EXPRESS MSE MESA AVIATION SERVICES AIR SHUTTLE EEC EXEC EXPRESS HUSTLER TAC THAI AIRLINES THAI-AIR DLH LUFTHANSA LUFTHANSA AWE AMERICAN WEST AIRLINES CACTUS NWA NORTHWEST AIRLINES NORTHWEST SWA SOUTHWEST AIRLINES SOUTHWEST
2. GENERAL AVIATION IDENTIFIERS
Aircraft are limited to a total of 7 alpha/numeric characters in the aircraft registration. In flight planning, Air Traffic Communications and automation data the letter N precedes the registration number to designate general aviation aircraft.
For brevity the caller/called section of the transcription should be reduced to N followed by the last 3 characters.
EXAMPLE: N45B (UNLESS THERE ARE NOT 3 CHARACTERS). Radio call for this aircraft would be November 45 Bravo, or the type aircraft may by substituted for the word November, e.g. Cessna 45 Bravo.
The minimum number of alpha/numeric characters is 2.
In event the aircraft is either an Air Taxi or Lifeguard flight it would be identified as follows:
AIR TAXI: TN45B Radio call would be Tango November 45 Bravo
LIFEGUARD: LN45B Radio call would be Lima November 45 Bravo
3. MILITARY SERVICE DESIGNATORS
Military aircraft are identified by a service designator followed by the aircraft's serial number.
DESIGNATOR SERVICE CALL SIGN/DESIGNATOR EXAMPLE ---------- ------- ---------------------------- A AIR FORCE A123456 VV NAVY VV3456 VM MARINE etc. R ARMY C COAST GUARD G AIR OR ARMY NATIONAL GUARDOther military aircraft use "tactical" call signs followed by numbers for A T C communications and flight plan information.
EXAMPLE: SPADE 42
4. NAVIGATION AIDS
The following navigation aids are frequently referred to in Air Traffic Communications.
VORTAC: V H F omni directional aid with distance measuring capability. Navigation aids are normally named after geographical locations near the facility, for example the BRIDGEPORT VORTAC is located near Bridgeport, Texas.
The three letter identifiers assigned to these facilities are provided by computer and are issued on a system wide basis to ensure that no two have the same identifier.
VORTAC NAME/3-LETTER IDENTIFIER ------------------------------- BRIDGEPORT/BPR BLUE RIDGE/BUJ SCURRY/SCY ACTON/AQN MILLSAP/MQP WACO/ACT ARDMORE/ADM5. INTERSECTIONS
Intersections are designated along airways formed by radials from VORTACS and are used for position determination and Air Traffic Control separation purposes. Intersections are assigned 5 letter designators by computer and are also issued system wide to ensure no duplication or confusion. In some instances the designators for intersections are named in correlation to geographical points. EXAMPLE: CELIN is near the town of Celina, Texas.
Intersections on Airway V66-278 between Bridgeport and Blue Ridge VORTACS, in sequence from west to east:
MYGAL JACKY KORKS CELIN KMART
Intersection on Airway V15 between Blue Ridge and Scurry VORTACs.
Intersections on Airway V16-94 between Scurry and Acton VORTACs, from east to west:
ELLMO RAINY BRITY KEEN MANDY
Intersection on Airway V17-163 between Scurry and Bridgeport VORTACs:
6. INTERSECTIONS ON STANDARD TERMINAL ARRIVAL ROUTES (STARS)
STARS serve as arrival routes into the DFW terminal area and are used for flight planning purposes and ATC clearances. Intersections on these routes are designated by 5 letter identifiers in the same manner as other intersections.
STAR NAME PRIMARY NAVIGATION AID RADIAL INTERSECTIONS ON ROUTE --------- ---------------------- ----- ---------------------- BOIDS 1 ARRIVAL BRIDGEPORT VORTAC BPR 108 RHOME BOIDS PIVIT ARINA BRIDGEPORT VORTAC BPR 115 CEVEE JURRO JUSTY HEAVN 3 ARRIVAL BRIDGEPORT VORTAC BPR 078 HEAVN MYGAL CELIN BLUE RIDGE 2 ARRIVAL BLUE RIDGE VORTAC BUJ 230 BRAID BATON ALKID HAMAK BUJ 215 DEANZ HOLTS WEDER SCURRY 9 ARRIVAL SCURRY VORTAC SCY 313 BECKX HERBO CAVEN SCY 296 PORKS SEAGO GLADD REFIL ACTON 1 ARRIVAL ACTON VORTAC AQN 040 MARKM BRYAR HULEN FLATO CREEK AQN 033 JERRY CRESN RENDY BOIDS 1 ARRIVAL BRIDGEPORT VORTAC BPR 108 RHOME BOIDS PIVIT ARINA BPR 115 CEVEE JURRO JUSTY HEAVN 3 ARRIVAL BRIDGEPORT VORTAC BPR 078 HEAVN MYGAL CELIN BLUE RIDGE 2 ARRIVAL BLUE RIDGE VORTAC BUJ 230 BRAID BATON ALKID HAMAK BUJ 215 DEANZ HOLTS WEDER SCURRY 9 ARRIVAL SCURRY VORTAC SCY 313 BECKX HERBO CAVEN SCY 296 PORKS SEAGO GLADD REFIL ACTON 1 ARRIVAL ACTON VORTAC AQN 040 MARKM BRYAR HULEN FLATO CREEK AQN 033 JERRY CRESN RENDY7. INTERSECTIONS AND FINAL APPROACH FIXES ON I L S SYSTEMS
ILS is an acronym for Instrument Landing System.
RUNWAYS: 18 RIGHT 18 LEFT 17 LEFT 17 RIGHT 13 LEFT -------- ------- ------- -------- ------- INTERSECTIONS: YOHAN ALIGN PENNY GARZA LEGRE UDALL ZINGG YALTA OUTER MARKERS: HASTY HASTY JIFFY JIFFY GUUDE RUNWAYS: 36 LEFT 35 RIGHT 31 RIGHT ------- -------- ------- INTERSECTIONS: CHARR DAYZZ VANDE HUTEN TUFFO GACHO OUTER MARKERS: BASIN ISSUE RIVER8. STANDARD INSTRUMENT DEPARTURES (SIDS)
SIDS are used by DFW, but since they are basically radar vector procedures to join a navigational radial, generally they are not called out by name. They are normally referred to by instructions for an aircraft to "JOIN THE SID".