GB2203878A - Speech recognition system - Google Patents
Speech recognition system Download PDFInfo
- Publication number
- GB2203878A GB2203878A GB8709322A GB8709322A GB2203878A GB 2203878 A GB2203878 A GB 2203878A GB 8709322 A GB8709322 A GB 8709322A GB 8709322 A GB8709322 A GB 8709322A GB 2203878 A GB2203878 A GB 2203878A
- Authority
- GB
- United Kingdom
- Prior art keywords
- word
- words
- speech recognition
- syntax tree
- node
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000010586 diagram Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000000034 method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
- G10L15/18—Speech classification or search using natural language modelling
- G10L15/183—Speech classification or search using natural language modelling using context dependencies, e.g. language models
- G10L15/19—Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
- G10L15/193—Formal grammars, e.g. finite state automata, context free grammars or word networks
Landscapes
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Machine Translation (AREA)
Abstract
A speech recognition system recognises sequences of words starting with one of a number of start words and defined by a predetermined syntax tree having a number of nodes from each of which extends at least one branch. Lock-up of the system due to the mis-recognition of a word is automatically released by arranging that the system returns to the start of a sequence whenever a start word is recognised, regardless of the node at which it occurs. The system may be used in an air traffic control training simulator. <IMAGE>
Description
SPEECH RECOGNITION SYSTEM
Speech recognition systems are designed to recognise certain words in human speech and to respond appropriately to a correct word or sequence of such words. Unless the system is intended to respond to single words only, then a predetermined pattern or sequence is usually established, called a syntax tree. This sets out the possible alternative sequences that the system has to recognise. For example the first word to be recognised may be one of several alternatives which may always be followed by one of several particular words. This in turn may be followed by one of several alternative words, and so on. There are no difficulties if the system recognises all words correctly. However, in practice, this does not always happen. Three types of error are common.The system may fail to recognise a word, in which case the system will cease to function correctly, becoming either locked at the current syntax node or possibly moving to the wrong syntax node after correct recognition of a subsequent word. This is known as rejection error. Alternatively the system may interpret one word as another word in the syntax tree, in which case it may follow an incorrect branch of the tree and be presented with a word. which does not follow in the sequence which the system is following.
This is known as substitution error. A third possibility, known as insertion error, is that a word which has not been used may be added, in which case the system may again follow the wrong branch in the syntax tree.
These problems have existed and been recognised for some time and the method of clearing such a block is to operate a switch which resets the system. This is an acceptable solution in many situations. However, if the speech recognition system is being used as some form of training aid and the resetting operation is one which the trainee would not normally perform, or if physical contact between the trainee and the system is impractical, then it is not an acceptable solution to the problem.
It is an object of the invention to- provide a speech recognition system having automatic means for resetting the system after faulty recognition of a word or words.
According to the present invention there is provided a speech recognition system arranged to recognise sequences of individual words starting with one of a number of start words and defined by a predetermined syntax tree having a series of nodes from each of which extends at least one branch, which system includes resetting means arranged such that, at each node, recognition of a start word returns the system to the beginning of a new sequence of words irrespective of the node in the syntax tree at which the start word is recognised.
The invention will now be described with reference to the accompanying drawings, in which
Figure 1 shows part of a syntax tree for an air traffic control training simulator; and
Figure 2 is a flow diagram illustrating the functioning of the system employing the syntax tree of Figure 1.
In the past it has been the practice to train air traffic controllers by arranging for voice communication between a trainee and a person who responds to instructions given by the trainee in the manner that an aircraft pilot would respond.
The person may be the instructor or an auxiliary member of the training team. Such an arrangement, whilst very simple technically, is very labour-intensive. Attempts have been made to provide training simulators in which a speech recognition system recognises words spoken by the trainee and controls a computer which operates a suitable response system. The problems already referred to may arise in such a situation.
Referring now to Figure 1, the syntax tree starts with the first word of any command sequence, referred to as a "start word" SW. This will be the first part of an aircraft callsign and may be the name used by air traffic controllers to designate aircraft of particular airlines. Four alternatives are shown, though of course there are more. Such a start word will always be followed at node 10 by a command word, initially the rest of the callsign, shown in Figure 1 as comprising three digits.
However, as shown in the drawing there are two branches extending from node 10, although the bnly word to be recognised should be one of the digits 0 to 9. The other branch contains all of the start words already mentioned. Recognition of the first digit leads to node 11 which again has two branches. One represents the second digit of the callsign whilst the other again represents all of the start words. A study of the syntax tree will show that at every node, where a command word is expected, there is an extra branch containing all of the start words.
Each start word branch leads back to the first node 10 in the syntax tree.
Clearly the speech recognition system may falter at any point. If, for example, the system does not recognise the first callsign digit spoken, then it will lock up.
Alternatively it may mis-recognise a word, for example interpreting "climb" as "turn". In such a situation it will expect the next word to be "left" or "right", whereas the trainee will say "to". The system will again lock up.
The invention, whilst not stopping the occurrence of lock-up, allows it to be automatically cleared when the trainee next speaks one of the start words. In real life, if a controllers call is not answered he will repeat it, starting with the callsign, and in the training situation this will reset the speech recognition system to node 10 in Figure 1.
The operation of the speech recognition system will be controlled by a program which defines the syntax tree. Figure 2 shows part of a flow diagram relating to such a program. At the beginning of the program the first step 20 is the recognition of a start word (SW). The next step, 21, requires the system to decide whether a recognised word is one of the next command words (CW) in the syntax tree or a start word. If the word is recognised as a command word then the program goes on to the next step 22. If the word is recognised as a start word then the system returns to the beginning of step 21. The third alternative is that the word is not recognised, in which case the system will lock up. However, in this situation the next recognised start word will reset the system to the beginning of step 21.
The same decision has to be made at each step in the syntax tree, as shown in Figure 2 until the last expected word in a sequence has been recognised.
Clearly the embodiment described of an air traffic control training simulator is only one application of the invention. The problems described affect all speech recognition systems and the invention may be applied to any system which uses a predetermined syntax tree.
Claims (3)
1. A speech recognition system arranged to recognise sequences of individual words starting with one of a number of start words and defined by a predetermined syntax tree having a number of nodes from each of which extends at least one branch, which system includes resetting means arranged such that, at each node, recognition of a start word returns the system to the beginning of a new sequence of words irrespective of the node in the syntax tree at which the start word is recognised.
2. A system as claimed in Claim 1 in which the system operates in accordance with a program having, at each word recognition step, a loop which returns the program to a predetermined point on recognition of a start word.
3. A speech recognition system substantially as herein described with reference to the accompanying drawings.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB8709322A GB2203878B (en) | 1987-04-21 | 1987-04-21 | Speech recognition system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB8709322A GB2203878B (en) | 1987-04-21 | 1987-04-21 | Speech recognition system |
Publications (3)
Publication Number | Publication Date |
---|---|
GB8709322D0 GB8709322D0 (en) | 1987-05-28 |
GB2203878A true GB2203878A (en) | 1988-10-26 |
GB2203878B GB2203878B (en) | 1991-04-17 |
Family
ID=10616051
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
GB8709322A Expired GB2203878B (en) | 1987-04-21 | 1987-04-21 | Speech recognition system |
Country Status (1)
Country | Link |
---|---|
GB (1) | GB2203878B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AT413809B (en) * | 1998-02-14 | 2006-06-15 | Tiefenbach Gmbh | Railway points control system |
-
1987
- 1987-04-21 GB GB8709322A patent/GB2203878B/en not_active Expired
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AT413809B (en) * | 1998-02-14 | 2006-06-15 | Tiefenbach Gmbh | Railway points control system |
Also Published As
Publication number | Publication date |
---|---|
GB2203878B (en) | 1991-04-17 |
GB8709322D0 (en) | 1987-05-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Schwanenflugel et al. | Interlingual semantic facilitation: Evidence for a common representational system in the bilingual lexicon | |
DE3853860D1 (en) | Device for producing a test-compatible, largely fault-tolerant configuration of redundantly implemented systolic VLSI systems. | |
EP0950966A3 (en) | Automatic routing method | |
GB2203878A (en) | Speech recognition system | |
Cohn | Literary technique in the Jeroboam narrative | |
Hirschman et al. | Beyond class A: A proposal for automatic evaluation of discourse | |
Okada | Complete inflation and perfect recall in extensive games | |
US3153121A (en) | Airport multi-language announcing system | |
US3155957A (en) | Signaling system | |
Dalmiya | Coherence, Truth and theOmniscient Interpreter | |
Rothwell | Recurrence Quantification Models of Human Conversational Grounding Processes: Informing Natural Language Human-Computer Interaction | |
Fry et al. | Automated tutoring dialogues for training in shipboard damage control | |
Shippey | The Curious Case of Denethor and the Palantír, Once More | |
SE455404B (en) | COMPUTER CONTROLS WITH DECENTRALIZED DEVICES, VALUE WITH TWO COMPUTERS | |
Gordon | Judgmental emotions | |
Gorfein et al. | When does “soar” become “sore”? Some comments on the chapter of Simpson and Kellas | |
Buckley | Temporal boundaries in Alsea | |
Dellner | The user’s role in automated fault detection and system recovery | |
EP0740448A3 (en) | Method for call blocking/unblocking in an ISDN telephone network | |
SU1429124A1 (en) | Device for modeling mass service systems | |
Hall | Colonial policy and neo-Melanesian | |
Schade et al. | Never Change a Winning Team: Towards Modelling Syntactic Priming in Language Production | |
Bacon | JE in the Middle Books of the Pentateuch. IV. Sinai-Horeb: Analysis of Exodus xviii.-xxxiv | |
Taylor | Vietnam: In the Beginning | |
Barch | Prediction of recurrent sequences as related to level of irrelevant cues. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PCNP | Patent ceased through non-payment of renewal fee |