Slot extraction for intents using large language models
Techniques for performing contextualized intent and slot extraction using a large language model (LLM) are disclosed. The LLM is generally pre-trained on an arbitrary corpus of language training data. A prompt is provided to the LLM. This prompt includes a limited number of prompt phrases. The prompt phrases share a semantic relationship with one another. A spoken utterance is recorded and then converted to text, resulting in generation of a transcription. The transcription is provided to the LLM. The LLM extracts, from the transcription, an extracted intent and an extracted slot. The extracted intent is determined to be related to a prompt-described intent that was included in the prompt. The prompt is supplemented by adding the extracted intent and the extracted slot to the prompt, resulting in the extracted intent being identified as sharing the semantic relationship with the other prompt phrases in the prompt.
1. A method for performing contextualized intent and slot extraction using a large language model (LLM), said method comprising:
accessing an LLM that is pre-trained on a corpus of language training data;
providing the LLM a prompt that includes a limited number of prompt phrases, wherein the prompt phrases share a semantic relationship with one another in that they correspond to a prompt-described intent, and wherein the prompt phrases use different vocabulary to describe the prompt-described intent;
providing a transcription of an utterance to the LLM;
causing the LLM to extract, from the transcription, an extracted intent and an extracted slot;
determining that the extracted intent is related to the prompt-described intent that was included in the prompt; and
supplementing the prompt by adding the extracted intent and the extracted slot to the prompt, resulting in the extracted intent being identified as sharing the semantic relationship with the other prompt phrases included in the prompt.
2. The method of claim 1 , wherein the utterance is provided within a context of an integrated development environment.
3. The method of claim 1 , wherein the corpus of language training data is an arbitrary corpus of language training data.
4. The method of claim 1 , wherein the corpus of language training data includes a plurality of different types of language inputs such that the LLM is trained using data that is different than utterances, intents, or slots.
5. The method of claim 1 , wherein a size of the prompt is limited to at most being a threshold size.
6. The method of claim 1 , wherein a size of the prompt is dependent on a determined complexity for the prompt-described intent.
7. The method of claim 1 , wherein the extracted intent is a previously unseen intent by the LLM.
8. The method of claim 1 , wherein the extracted slot is a previously unseen slot by the LLM.
9. The method of claim 1 , wherein the LLM determines that the utterance is received within a context of an integrated development environment (IDE), and wherein the LLM tailors an output based on the determined context.
10. The method of claim 1 , wherein the extracted intent is a predicted intent that is determined by the LLM based on a determined context associated with the utterance.
11. The method of claim 1 , wherein the transcription of the utterance does not match a previous transcription for a previous utterance that is known to the LLM.
12. The method of claim 1 , wherein a first prompt phrase included among said prompt phrases indicates what portion of the first prompt phrase constitutes a slot in that first prompt phrase.
13. The method of claim 1 , wherein the prompt is a text-based file.
14. The method of claim 1 , wherein the limited number of prompt phrases included in the prompt is less than 10 prompt phrases.
15. The method of claim 14 , wherein the limited number of prompt phrases included in the prompt is less than 5 prompt phrases.
16. A computer system that performs contextualized intent and slot extraction using a large language model (LLM), said computer system comprising:
at least one processor; and
at least one hardware storage device that stores instructions that are executable by the at least one processor to cause the computer system to:
access an LLM that is pre-trained on an arbitrary corpus of language training data;
provide the LLM a prompt that includes a limited number of prompt phrases, wherein the prompt phrases share a semantic relationship with one another in that they correspond to a prompt-described intent, and wherein the prompt phrases use different vocabulary to describe the prompt-described intent;
access a transcription of an utterance;
provide the transcription to the LLM;
cause the LLM to extract, from the transcription, an extracted intent and an extracted slot;
determine that the extracted intent is related to the prompt-described intent that was included in the prompt; and
supplement the prompt by adding the extracted intent and the extracted slot to the prompt, resulting in the extracted intent being identified as sharing the semantic relationship with the other prompt phrases included in the prompt.
17. The computer system of claim 16 , wherein the utterance is a programming command spoken within a context of an integrated development environment (IDE).
18. The computer system of claim 16 , wherein the prompt operates as a running log that records various alternative techniques for triggering the prompt-described intent.
19. The computer system of claim 16 , wherein the utterance is received in real time, and the transcription is generated in real time.
20. The computer system of claim 16 , wherein the limited number of prompt phrases included in the prompt is less than 20 prompt phrases.
21. A method for performing contextualized intent and slot extraction using a large language model (LLM), said method comprising: accessing an LLM that is pre-trained on an arbitrary corpus of language training data; providing the LLM a prompt that includes a limited number of prompt phrases, wherein the prompt phrases share a semantic relationship with one another in that they correspond to a prompt-described intent, and wherein the prompt phrases use different vocabulary to describe the prompt-described intent; accessing a transcription of an utterance; providing the transcription to the LLM; causing the LLM to extract, from the transcription, an extracted intent and an extracted slot; determining that the extracted intent is related to the prompt-described intent that was included in the prompt; and supplementing the prompt by adding the extracted intent and the extracted slot to the prompt, resulting in the extracted intent being identified as sharing the semantic relationship with the other prompt phrases included in the prompt.