Autonomous content orchestration
A system could include persistent storage units. Each persistent storage unit could contain: (i) records including descriptions related to technology-related problems, and (ii) term frequency (TF) metrics for tokens within the descriptions. The system could include processors configured to perform operations. The operations may include receiving input text and determining that the input text is subject to a querying procedure. The operations could also include, in response to the determining, identifying a classification for the input text and based on the classification, selecting at least one of the persistent storage units. The operations could further include tokenizing the input text to form input tokens and calculating, for the input tokens, input TF metrics. The operations could additionally include determining similarity metrics between: (i) the input TF metrics, and (ii) TF metrics for records within the at least one persistent storage unit and, based on the similarity metrics, retrieving relevant records.
1. A system comprising:
one or more persistent storage units, each persistent storage unit containing: (i) a plurality of records, wherein the plurality of records include text-based descriptions related to technology-related problems experienced by users, and (ii) term frequency metrics for text-based tokens identified within the text-based descriptions; and
one or more processors configured to perform operations, including:
receiving, from a source device, input text;
determining that the input text is subject to a querying procedure;
in response to determining that the input text is subject to the querying procedure, identifying a classification for the input text;
based at least on the classification, selecting at least one of the one or more persistent storage units;
tokenizing the input text to form a set of input text-based tokens for the input text;
calculating, for the set of input text-based tokens, input term frequency metrics;
determining similarity metrics between: (i) the input term frequency metrics, and (ii) term frequency metrics for records disposed within the at least one selected persistent storage unit;
based on the similarity metrics, retrieving, from the at least one selected persistent storage unit, a relevant set of the records; and
providing, to the source device and from the relevant set of the records, text-based descriptions.
2. The system of claim 1 , wherein the operations further include:
receiving, from a second source device, second input text;
determining that the second input text is not subject to the querying procedure; and
in response determining that the second input text is not subject to the querying procedure, storing, as a record in at least one of the one or more persistent storage units, the second input text and second input term frequency metrics for text-based tokens identified within the second input text.
3. The system of claim 1 , wherein both the text-based tokens identified within the text-based descriptions and the input text-based tokens lack stop words, common words, and punctuation, wherein the stop words comprise a pre-defined list of words related to generic technical topics, and wherein the common words comprise a pre-defined list of common English language words.
4. The system of claim 1 , wherein determining that the input text is subject to the querying procedure comprises:
identifying, based on textual content of the input text, a textual sentiment metric;
identifying, based on the textual content of the input text, a verbosity metric;
providing, to a trained machine learning (ML) model, the textual sentiment metric and the verbosity metric, wherein the trained ML model is configured to receive metrics related to text and output a prediction of whether the text is subject to querying; and
receiving, from the trained ML model, a prediction that the input text is subject to querying.
5. The system of claim 4 , wherein the trained ML model is configured to output an accuracy measurement indicating a confidence in the prediction of whether the text is subject to querying, and wherein determining that the input text is subject to the querying procedure comprises:
determining that an accuracy measurement for the prediction that the input text is subject to querying is higher than a pre-defined threshold accuracy.
6. The system of claim 1 , wherein determining that the input text is subject to the querying procedure comprises:
identifying a particular web interface through which the input text was provided, and
based on the particular web interface, determining that the input text is subject to the querying procedure.
7. The system of claim 1 , wherein a particular term frequency metric is for a text-based token in a text-based description of a particular record in the plurality of records, and wherein the particular term frequency metric is based on a product of: (i) a number of times that the text-based token appears in the text-based description, and (ii) a logarithm of a quotient of a count of the plurality of records and a count of records from the plurality of records that contain the text-based token.
8. The system of claim 1 , wherein the relevant set of the records is a sequence of n most similar records ranked based on the similarity metrics.
9. The system of claim 1 , wherein retrieving the relevant set of the records comprises filtering out records that do not match a pre-determined search scope, wherein the pre-determined search scope controls an extent to which, for each relevant record of the relevant set of the records, each of the set of input text-based tokens is required to be in a text-based description of the relevant record.
10. The system of claim 9 , wherein the pre-determined search scope is based on a calculation, for each particular record in the relevant set of the records, based on a quotient of (i) an intersection of the set of input text-based tokens and a tokenized versions of the text-based description for the particular record, and (ii) a count of the set of input text-based tokens.
11. The system of claim 10 , wherein the input text is entered into a graphical user interface of the source device, wherein a representation of the graphical user interface was generated and transmitted to the source device by the one or more processors, and wherein the graphical user interface contains a drop-down menu that controls a threshold search scope value, and wherein filtering out records that do not match the pre-determined search scope comprises:
filtering out records with quotients that are below the threshold search scope value.
12. The system of claim 10 , wherein the operations further comprise setting, based on the classification, a threshold search scope value, and wherein filtering out records that do not match the pre-determined search scope comprises filtering out records with quotients that are below the threshold search scope value.
13. The system of claim 1 , wherein each persistent storage unit further contains usefulness metrics respectively corresponding to the text-based descriptions, and wherein retrieving the relevant set of the records comprises filtering out records with text-based descriptions that correspond to usefulness metrics outside of a pre-determined range.
14. The system of claim 13 , wherein a particular usefulness metric is for a tokenized version of a particular text-based description, and wherein the particular usefulness metric is based on a quotient of: (i) a first count of tokens in the tokenized version of the particular text-based description that lacks stop words, common words, and punctuation, and (ii) a second count of tokens in the tokenized version of the particular text-based description that includes the stop words, common words, and punctuation, wherein the stop words comprise a pre-defined list of words related to generic technical topics, and wherein the common words comprise a pre-defined list of common English language words.
15. The system of claim 13 , wherein the pre-determined range is based on the classification.
16. A computer-implemented method comprising:
receiving, by one or more processors disposed within a system and from a source device, input text, wherein the system includes one or more persistent storage units, each persistent storage unit containing: (i) a plurality of records, wherein the plurality of records include text-based descriptions related to technology-related problems experienced by users, and (ii) term frequency metrics for text-based tokens identified within the text-based descriptions;
determining, by the one or more processors, that the input text is subject to a querying procedure;
in response to determining that the input text is subject to the querying procedure, identifying, by the one or more processors, a classification for the input text;
based at least on the classification, selecting, by the one or more processors, at least one of the one or more persistent storage units;
tokenizing, by the one or more processors, the input text to form a set of input text-based tokens for the input text;
calculating, by the one or more processors and for the set of input text-based tokens, input term frequency metrics;
determining, by the one or more processors, similarity metrics between: (i) the input term frequency metrics, and (ii) term frequency metrics for records disposed within the at least one selected persistent storage unit;
based on the similarity metrics, retrieving, by the one or more processors and from the at least one selected persistent storage unit, a relevant set of the records; and
from the relevant set of the records, providing, by the one or more processors and to the source device, text-based descriptions.
17. The computer-implemented method of claim 16 , further comprising:
receiving, from a second source device, second input text;
determining that the second input text is not subject to the querying procedure; and
in response determining that the second input text is not subject to the querying procedure, storing, as a record in at least one of the one or more persistent storage units, the second input text and second input term frequency metrics for text-based tokens identified within the second input text.
18. The computer-implemented method of claim 16 , wherein determining that the input text is subject to the querying procedure comprises:
identifying, based on textual content of the input text, a textual sentiment metric;
identifying, based on the textual content of the input text, a verbosity metric;
providing, to a trained machine learning (ML) model, the textual sentiment metric and the verbosity metric, wherein the trained ML model is configured to receive metrics related to text and output a prediction of whether the text is subject to querying; and
receiving, from the trained ML model, a prediction that the input text is subject to querying.
19. The computer-implemented method of claim 16 , wherein a particular term frequency metric is for a text-based token in a text-based description of a particular record in the plurality of records, and wherein the particular term frequency metric is based on a product of: (i) a number of times that the text-based token appears in the text-based description, and (ii) a logarithm of a quotient of a count of the plurality of records and a count of records from the plurality of records that contain the text-based token.
20. An article of manufacture including a non-transitory computer-readable medium, having stored thereon program instructions that, upon execution by a computing system, cause the computing system to perform operations comprising:
receiving, from a source device, input text, wherein the computing system includes one or more persistent storage units, each persistent storage unit containing: (i) a plurality of records, wherein the plurality of records include text-based descriptions related to technology-related problems experienced by users, and (ii) term frequency metrics for text-based tokens identified within the text-based descriptions;
determining that the input text is subject to a querying procedure;
in response to determining that the input text is subject to the querying procedure, identifying a classification for the input text;
based at least on the classification, selecting at least one of the one or more persistent storage units;
tokenizing the input text to form a set of input text-based tokens for the input text;
calculating, for the set of input text-based tokens, input term frequency metrics;
determining similarity metrics between: (i) the input term frequency metrics, and (ii) term frequency metrics for records disposed within the at least one selected persistent storage unit;
based on the similarity metrics, retrieving, from the at least one selected persistent storage unit, a relevant set of the records; and
providing, from the relevant set of the records and to the source device, text-based descriptions.