CodePudding user response:
https://aws.amazon.com/polly/But at present does not support Chinese,
I guess the possible is
Call API to achieve recognition
http://www.iflytek.com/Then compare the text problem, the answer to the last play recorded
CodePudding user response:
General idea is to use a context of dialogue scenes, to determine the content of the dialogue below, such as people ask me where I can find a restaurant nearby? The machine can try to ask what would you like to eat? Man if you want to eat roast duck, then the context has been transferred to eat roast duck the background, and then continue to ask consumption level has a limit? If they answer no, or you want to no more than 100 yuan, and then turn to the corresponding context, thus becomes a state transition diagram,CodePudding user response:
First, the infection worm system structure of database,Second, sound collection and distribution, conversion database,
Third, a searchable database of fuzzy judgment analysis subsystem,
Fourth, hardware matching, retrieval, database calls,
Fifth, analysis of formula, the data simulation formula, data analysis formula framework, content,
Sixth, chat room structure components, and chat logs backup,
CodePudding user response:
Need a lot of dialogue as the answer sample, such as phone service companies are collecting vast amounts of phone call records, as the interaction sampleCodePudding user response:
Do a little bit more simple, that is, the semantic analysis:Such as watching dialogue:
-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
Ask: "did you eat?"
Answer a: "eat"
Q 2: "what's the weather like today?"
Answer two: "clear"
-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
Need to extract the keywords, ask a keyword is to have a meal, ask
Ask two keywords is "today, the weather
"