System and method for ai-assisted relationship journaling and sentiment analysis

A method for implementing a psychological model for relationship therapy, comprising:

• enrolling a licensed therapist by:
• receiving therapist registration information via a web-based interface;
• processing subscription payment for annual access to a therapy platform;
• creating a therapist account record in a relational database with unique authentication credentials;
• and
• granting the therapist access to a reporting dashboard;
• registering a client pair by:
• receiving identification information for a first client and a second client from the therapist;
• generating unique login credentials for the first client and the second client;
• creating client records in the relational database linked to the therapist account record;
• and
• providing mobile application installation instructions and login credentials to the first client and the second client;
• receiving journal entries by:
• accepting text input via a mobile application interface executing on a client mobile device, wherein the text input describes a relationship interaction between the first client and the second client;
• transmitting the text input to a sentiment analysis engine hosted on a remote server;
• analyzing sentiment by:
• parsing the text input into semantic units using natural language processing;
• identifying emotional valence of each semantic unit based on a lexicon of relationship-relevant terms;
• classifying the text input as representing care, neglect, or abuse based on the identified emotional valence;
• assigning a numerical score on a scale ranging from −10 to +10, wherein positive scores indicate care, zero indicates neglect, and negative scores indicate abuse;
• detecting presence of repair attempt language within the text input;
• applying a +2 modifier to the numerical score when repair attempt language is detected;
• generating a confidence score indicating certainty of the sentiment classification;
• and
• transmitting the numerical score, confidence score, and reasoning explanation back to the mobile application;
• storing event data by:
• creating an event record in the relational database comprising the numerical score, a timestamp, and metadata identifying the client pair;
• maintaining the full text of the journal entry locally on the client mobile device without transmitting the full text to the remote server;
• generating reports by:
• receiving a request from the therapist via the reporting dashboard to view sentiment data for the client pair;
• querying the relational database for event records associated with the client pair within a specified date range;
• calculating summary statistics comprising counts of neglect events, repair attempt events, shared entries and bid attempts;
• generating a scatter plot visualization displaying the numerical scores plotted chronologically and color coded;
• green for positive scores (Care), otherwise red;
• overlaying a trend line on the scatter plot using linear regression or moving average calculation with positive slope color coded green, otherwise red;
• and
• presenting the scatter plot and summary statistics to the therapist via the reporting dashboard.

A non-transitory computer-readable storage medium containing instructions that, when executed by a processor, cause the processor to implement a Care-Neglect-Abuse continuum psychological model by:

• receiving journal entry text via an application programming interface (API),
• wherein the journal entry text comprises a narrative description of a relationship interaction;
• tokenizing the journal entry text into a plurality of words and phrases;
• removing stop words from the plurality of words and phrases to generate a set of semantically significant terms;
• comparing the set of semantically significant terms against a sentiment lexicon comprising:
• care-related terms associated with positive numerical values ranging from +1 to +10;
• neglect-related terms associated with a zero numerical value;
• and
• abuse-related terms associated with negative numerical values ranging from −1 to −10;
• calculating a preliminary sentiment score by aggregating numerical values associated with matching terms from the sentiment lexicon;
• analyzing contextual information surrounding the matching terms to adjust the preliminary sentiment score, wherein the contextual information comprises:
• negation words that reverse the polarity of adjacent terms;
• intensifier words that amplify the magnitude of adjacent terms;
• and
• emotion words that indicate subjective emotional states;
• searching the journal entry text for repair attempt indicators comprising:
• apology language patterns;
• reconciliation effort phrases;
• responsibility-taking statements;
• and
• soothing behavior expressions;
• when one or more repair attempt indicators are detected:
• adding a +2 modifier to the preliminary sentiment score to generate a final sentiment score;
• and
• setting a repair attempt flag to true;
• when no repair attempt indicators are detected:
• designating the preliminary sentiment score as the final sentiment score;
• and
• setting the repair attempt flag to false;
• calculating a confidence score ranging from 0.0 to 1.0 based on:
• strength of lexicon matches;
• clarity of emotional language;
• and
• absence of ambiguous or sarcastic phrasing;
• generating a reasoning explanation describing:
• which terms contributed to the final sentiment score;
• whether care, neglect, or abuse was detected;
• and
• whether a repair attempt was identified;
• transmitting a response comprising the final sentiment score, the repair attempt flag, the confidence score, and the reasoning explanation via the application programming interface;
• and
• storing the final sentiment score, the repair attempt flag, bid, shared, and neglect flags in a database record associated with a client identifier and a timestamp, wherein the journal entry text is not stored in the database record.