Techniques for joint context query rewrite and intent detection
Artificial intelligence techniques for query management are described. A method comprises generating, by a context detection module, context information for a first query comprising natural language information to request a result from one of a plurality of machine learning models, modifying, by a query modification module, the first query based the context information to form a first modified query, determining, by an intent module, an intent type for the first modified query, selecting, by a routing module, a machine learning model from the plurality of machine learning models based on the intent type, and routing, by the routing module, the first modified query to the selected machine learning model. Other embodiments are described and claimed.
1 . A system comprising:
a memory component; and
one or more processing devices coupled to the memory component, the one or more processing devices to perform operations comprising:
generating, by a context detection module, context information for a first query of a query session during a first iterative process, the first query comprising multimodal information to request a result from one of a plurality of machine learning models;
modifying, by a query modification module, the first query based on the context information to form a first modified query;
determining, by an intent module, an intent type for the first modified query from multiple intent types generated by a combination of a machine learning model and rule-based logic using the first modified query, the machine learning model trained using one-shot or few-shot learning training samples, and wherein the intent type is an output from the machine learning model that matches an output from the rule-based logic;
selecting, by a routing module, a machine learning model from the plurality of machine learning models based on the intent type; and
routing, by the routing module, the first modified query to the selected machine learning model.
2 . The system of claim 1 , the one or more processing devices to perform operations comprising extracting, by a context extraction module of the context detection module, query context information comprising context information from the first query.
3 . The system of claim 1 , the one or more processing devices to perform operations comprising:
generating, by the context detection module, context information for a second query of the query session and the first modified query during a second iterative process, the second query comprising multimodal information to request a result from one of the plurality of machine learning models;
modifying, by the query modification module, the second query based the context information to form a second modified query;
determining, by an intent module, an intent type for the second modified query;
selecting, by a routing module, a machine learning model from the plurality of machine learning models based on the intent type; and
routing, by the routing module, the second modified query to the selected machine learning model.
4 . The system of claim 3 , the one or more processing devices to perform operations comprising extracting, by a context extraction module of the context detection module, query context information and modified query context information, the query context information comprising context information from the second query and the modified query context information comprising context information from the first modified query.
5 . The system of claim 1 , the one or more processing devices to perform operations comprising determining, by an intent inference model of the intent module, the intent type for the first modified query, wherein the intent inference model is the machine learning model trained to predict different intent types.
6 . The system of claim 1 , the one or more processing devices to perform operations comprising determining, by an intent detector module of the intent module, the intent type of the first modified query, wherein the intent detector module uses a set of intent definitions corresponding to different intent types.
7 . The system of claim 1 , the one or more processing devices to perform operations comprising determining, by an intent inference model and an intent detector module of the intent module, the intent type for the first modified query, wherein the intent inference model and the intent detector module operate in parallel.
8 . The system of claim 1 , the one or more processing devices to perform operations comprising determining, by an intent inference model and an intent detector module of the intent module, the intent type for the first modified query, wherein the intent inference model and the intent detector module operate in sequence.
9 . A method, comprising:
generating, by a context detection module, query context information for a first query of a query session during a first iterative process, the first query comprising multimodal information to request a result from a first machine learning model;
modifying, by a query modification module, the first query based on the query context information for the first query to form a first modified query;
generating, by the context detection module, query context information for a second query of the query session and modified query context information for the first modified query during a second iterative process, the second query comprising multimodal information to request a result from a second machine learning model;
modifying, by the query modification module, the second query based on the query context information for the second query and the modified query context information for the first modified query to form a second modified query, wherein the second modified query comprises a recursive summary of the query context information for the first query, the modified query context information for the first modified query, and the query context information for the second query; and
routing, by a routing module, the first modified query to the first machine learning model and the second modified query to the second machine learning model.
10 . The method of claim 9 , wherein the multimodal information comprises natural language text information, further comprising extracting, by a context extraction module of the context detection module, the query context information from natural language text information of the first query during the first iterative process.
11 . The method of claim 9 , wherein the multimodal information comprises natural language text information, further comprising extracting, by a context extraction module of the context detection module, the query context information from natural language text information of the second query and the modified query context information from natural language text information of the first modified query during the second iterative process.
12 . A system comprising:
a memory component; and
one or more processing devices coupled to the memory component, the one or more processing devices to perform operations comprising:
generating, by a context detection module, context information for a query comprising natural language information to request a result from one of a plurality of machine learning models;
determining, by an intent module, an intent type for the query from multiple intent types generated by a combination of a machine learning model and rule-based logic, the machine learning model trained using one-shot or few-shot learning training samples, and wherein the intent type is an output from the machine learning model that matches an output from the rule-based logic;
selecting, by a routing module, a machine learning model from the plurality of machine learning models based on the intent type; and
routing, by the routing module, the query to the selected machine learning model.
13 . The system of claim 12 , the one or more processing devices to perform operations comprising determining, by an intent inference model of the intent module, the intent type for the query, wherein the intent inference model is the machine learning model trained to predict different intent types.
14 . The system of claim 12 , the one or more processing devices to perform operations comprising determining, by an intent detector module of the intent module, the intent type of the query, wherein the intent detector module uses a set of intent definitions corresponding to different intent types.
15 . The system of claim 12 , the one or more processing devices to perform operations comprising:
determining, by an intent inference model of the intent module, a first intent type for the query;
determining, by an intent detector module of the intent module, a second intent type for the query;
comparing, by the intent module, the first intent type and the second intent type; and
determining, by the intent module, the intent type for the query when the first intent type matches the second intent type.
16 . The system of claim 12 , the one or more processing devices to perform operations comprising determining, by an intent inference model and an intent detector module of the intent module, the intent type for the query, wherein the intent inference model and the intent detector module operate in parallel.
17 . The system of claim 12 , the one or more processing devices to perform operations comprising determining, by an intent inference model and an intent detector module of the intent module, the intent type for the query, wherein the intent inference model and the intent detector module operate in sequence.
18 . The system of claim 12 , wherein the intent type comprises visualization, forecasting, anomaly detection, data question and answer, breakdown dimension, segment creation, or summary captioning.