Generating action items during a conferencing session
Embodiments include systems and methods for receiving an action item trigger by a user of a conferencing application; and in response to receiving the action item trigger, generating spoken words from audio data of a session of the conferencing application; normalizing the spoken words; generating higher-level representations of the normalized spoken words; determining semantic similarities of the higher-level representations of the normalized spoken words and higher level representations of normalized action words of an action word list; ranking options for top spoken words and action words based at least in part on the semantic similarities; identifying candidates for action words and/or phrases from the top spoken words and action words; and parsing the candidates to generate one or more action items.
1. A method comprising:
receiving an action item trigger by a user of a conferencing application; and
in response to receiving the action item trigger:
generating spoken words from audio data of a session of the conferencing application for a previous predetermined number of sentences spoken before receiving the action item trigger;
normalizing the spoken words;
generating higher level representations of the normalized spoken words;
determining semantic similarities of the higher-level representations of the normalized spoken words and higher-level representations of normalized action words of an action word list;
ranking options for top spoken words and action words based at least in part on the semantic similarities;
identifying candidates for action words and/or phrases from the top spoken words and action words; and
parsing the candidates to generate one or more action items.
2. The method of claim 1 , comprising normalizing action words in the action word list and generating higher-level representations of the normalized action words.
3. The method of claim 1 , comprising parsing the candidates to generate one or more action items by:
dividing, by a trained neural network, candidate action phrases into words based at least in part on data from a dependency databank;
classifying the words into dependency classes according to a set of rules; and
generating one or more action items based at least in part on the dependency classes.
4. The method of claim 1 , wherein the action item trigger comprises an input selection in real-time by the user of a user interface of the conferencing application.
5. The method of claim 1 , wherein the one or more action items comprises a task assigned to be performed by a person based on analysis of the audio data of the session of the conferencing application.
6. The method of claim 5 , comprising sending an email to the person's email account notifying the person of the assigned task.
7. The method of claim 1 , comprising generating spoken words from audio data of the session of the conferencing application for a previous predetermined period of time before receiving the action item trigger instead of for the previous predetermined number of sentences spoken before receiving the action trigger.
8. The method of claim 1 , comprising excluding spoken words attributable to the user.
9. The method of claim 1 , wherein the action word list comprises a private lexicon.
10. At least one non-transitory machine-readable storage medium comprising instructions that, when executed, cause at least one processor to:
receive an action item trigger by a user of a conferencing application; and
in response to receiving the action item trigger:
generate spoken words from audio data of a session of the conferencing application for a previous predetermined number of sentences spoken before receiving the action item trigger;
normalize the spoken words;
generate higher-level representations of the normalized spoken words;
determine semantic similarities of the higher-level representations of the normalized spoken words and higher-level representations of normalized action words of an action word list;
rank options for top spoken words and action words based at least in part on the semantic similarities;
identify candidates for action words and/or phrases from the top spoken words and action words; and
parse the candidates to generate one or more action items.
11. The least one non-transitory machine-readable storage medium of claim 10 comprising instructions that, when executed, cause at least one processor to normalize action words in the action word list and generate higher level representations of the normalized action words.
12. The least one non-transitory machine-readable storage medium of claim 10 comprising instructions that, when executed, cause at least one processor to parse the candidates to generate one or more action items by:
dividing, by a trained neural network, candidate action phrases into words based at least in part on data from a dependency databank;
classifying the words into dependency classes according to a set of rules; and
generating one or more action items based at least in part on the dependency classes.
13. The least one non-transitory machine-readable storage medium of claim 10 , wherein the action item trigger comprises an input selection in real-time by the user of a user interface of the conferencing application.
14. The least one non-transitory machine-readable storage medium of claim 10 , wherein the one or more action items comprises a task assigned to be performed by a person based on analysis of the audio data of the session of the conferencing application.
15. An apparatus comprising:
a conferencing manager to receive audio data and an action item trigger; and
an action item generator to receive an action item trigger by a user of a conferencing application; and in response to receiving the action item trigger, to generate spoken words from the audio data of a session of the conferencing application for a previous predetermined number of sentences spoken before receiving the action item trigger; normalize the spoken words; generate higher level representations of the normalized spoken words; determine semantic similarities of the higher level representations of the normalized spoken words and higher level representations of normalized action words of an action word list; rank options for top spoken words and action words based at least in part on the semantic similarities; identify candidates for action words and/or phrases from the top spoken words and action words; and parse the candidates to generate one or more action items.
16. The apparatus of claim 15 wherein the action item generator is to normalize action words in the action word list and generate higher level representations of the normalized action words.
17. The apparatus of claim 15 wherein the action item generator to parse the candidates to generate one or more action items by dividing, by a trained neural network, candidate action phrases into words based at least in part on data from a dependency databank; classifying the words into dependency classes according to a set of rules; and generating one or more action items based at least in part on the dependency classes.
18. The apparatus of claim 15 , wherein the action item trigger comprises an input selection in real-time by the user of a user interface of the conferencing application.
19. The apparatus of claim 15 , wherein the one or more action items comprises a task assigned to be performed by a person based on analysis of the audio data of the session of the conferencing application.