One of the major challenges of modern political campaigning is to identify and address the predominant issues that a constituency is focused on. Telephone surveys and focus groups are expensive, restricted in scope and are pre-structured to quantify sentiment towards known issues and contenders, with limited capability to identify the “dark horses” of politics, the “unknown unknowns” that may have a critical effect on the political scene and local issues of attention of particular communities that can be leveraged to swing the vote in those sectors. The ability of a campaign to tailor messages to different communities and social sectors and to project a sense that the candidate knows what is "on their minds", is "talking to them" and "listening to them" can have a decisive influence.
In order to meet this need, IntuView offers its technology to analyse diverse textual sources and to enable the user to identify:
- “Who is talking” - who are generating the chatter (sociologically, ethnically, age-wise based on their social frames of reference in the language.
- “How many” (the scope of the situation that you want to be aware of).
- “What are they are talking about” (entity recognition, disambiguation and resolution, including “idea entities”)
- “Where they are” - not only according to meta-data but by content,
- “Who are the talking about” – who is the object of the chatter.
- “How and why do they feel about the object” - sentiment (not only “like” or “dislike”, but the precise sentiment and the reason)
IntuScan can receive a random stream of information from different sources pre-identified with the relevant geographical region, as follows:
1. Social media that can be accessed (tweets, blogs associated geographically with the area of focus)
2. Local news (online newspapers etc.)
3. Directed Emails/tweets for specific events initiated by the campaign ("town house" meetings etc.)
Alternatively, the system may be set up to perform real-time analysis of feedback initiated by the campaign itself.
This can be implemented as follows:
- Prior to a meeting, the organizers can reach out to potential viewers through emails, text messages and phone.
- In this communication they can offer the constituents an opportunity to have their voices heard by expressing their opinions and questions to the candidate prior to the event by Email, a website form, Twitter or SMS.
- The system will then analyze the communications and present a list of issues with corresponding information regarding what sector of the constituents focused on those issues and the terminology used.
The output of the system will provide identification of the following elements:
1. The social strata in terms of age, ethnicity, professional background of the texts in order to identify what issues are in the focus of which age/social groups.
2. Entities (persons, organizations, places, events) - the different occurrences of the entities – even in different languages and orthographies - will be aggregated to reference to any entity in all its permutations such as: with or without a title, using an oblique reference (the Representative from...), positive or negative nicknames (Camelot, “Tricky Dicky”, The Gipper, “The Messiah”) etc.
3. Issues as expressed in different ways that identify the political inclinations of the authors. This will include ideas that were not pre-defined that appear in the information stream and may indicate “unknown unknowns”
4. Sentiment towards specific entities and, indirectly, towards their “parent entities”.
The output will be inserted into a database with a user interface that enables the user to view clusters of references such as “the objects towards which people of a certain sector express the most positive/negative sentiment” or “the issues that are most common in the discourse in a certain city”. In addition, the platform can provide real-time feedback on the responses of the same feeds (twitter etc.) to televised or online events. This feature can be used to support "townhouse meetings" of political campaigns and to assess impact of candidates in televised appearances. An enhanced capability can use the IntuView ontological categorization tools. For such an implementation, IntuView can use batches of information that have been pre-assessed as relating to different issues as training sets for prediction of the categories of input documents.