The Whole Truth and Nothing But the Truth: Faithful and Controllable Dialogue Response Generation...
646
16.6
Microsoft Research320 тыс
Следующее
299 дней – 77049:01Популярные
01.12.22 – 48415:52
08.02.22 – 70530:50Опубликовано 7 июля 2023, 18:47
The Whole Truth and Nothing But the Truth: Faithful and Controllable Dialogue Response Generation with Dataflow Transduction and Constrained Decoding
In a real-world dialogue system, generated text must be truthful and informative while remaining fluent and adhering to a prescribed style. Satisfying these constraints simultaneously is difficult for the two predominant paradigms in language generation: neural language modeling and rule-based generation. We describe a hybrid architecture for dialogue response generation that combines the strengths of both paradigms. The first component of this architecture is a rule-based content selection model defined using a new formal framework called dataflow transduction, which uses declarative rules to transduce a dialogue agent's actions and their results (represented as dataflow graphs) into context-free grammars representing the space of contextually acceptable responses. The second component is a constrained decoding procedure that uses these grammars to constrain the output of a neural language model, which selects fluent utterances. Our experiments show that this system outperforms both rule-based and learned approaches in human evaluations of fluency, relevance, and truthfulness.
See more at microsoft.com/en-us/research/v...
In a real-world dialogue system, generated text must be truthful and informative while remaining fluent and adhering to a prescribed style. Satisfying these constraints simultaneously is difficult for the two predominant paradigms in language generation: neural language modeling and rule-based generation. We describe a hybrid architecture for dialogue response generation that combines the strengths of both paradigms. The first component of this architecture is a rule-based content selection model defined using a new formal framework called dataflow transduction, which uses declarative rules to transduce a dialogue agent's actions and their results (represented as dataflow graphs) into context-free grammars representing the space of contextually acceptable responses. The second component is a constrained decoding procedure that uses these grammars to constrain the output of a neural language model, which selects fluent utterances. Our experiments show that this system outperforms both rule-based and learned approaches in human evaluations of fluency, relevance, and truthfulness.
See more at microsoft.com/en-us/research/v...
Свежие видео
1 день – 92 51119:14
2 дня – 4840:59
5 дней – 4 8080:20
5 дней – 5712:33
5 дней – 1 45010:19
6 дней – 1 0340:45Случайные видео
52 дня – 9411:59
58 дней – 21 00250:34
28.01.22 – 1 7021:46
24.09.20 – 1 767 41618:34
21.05.16 – 50 4373:55
1 день – 9120:11
2 дня – 1 1980:44
2 дня – 3472:53
2 дня – 514:10
3 дня – 5 2330:52
6 дней – 2 28732:25
7 дней – 7861:10
11 часов – 2 0110:31