[BUG] LexC Document Parsing Failures: Document Type, Entity Extraction, and Party Identification

Bug Description

Multiple critical parsing failures identified in LexC document processing that affect core entity extraction, document classification, and metadata generation.

Test Document Analysis

Using a standard consulting agreement document ("SchromM_Services Agreement_20190829.docx"), LexC demonstrates systematic parsing failures across multiple categories:

1. Document Type Misidentification

• Expected: "contract" or "consulting_agreement"
• Actual: Type: unknown
• Evidence: Document header clearly states "CONSULTING AGREEMENT"
• Impact: Fundamental document classification failure

2. Party Identification Failures

• Jennifer Liotta: Incorrectly labeled as [ORGANIZATION:0.90:Client] instead of PERSON/PARTY
• Micha Schrom: Correctly identified as [PERSON:0.50:Micha Schrom] but with suspicious low confidence (0.50)
• Metadata shows: parties: [] (empty array) despite two clearly named parties

3. Missing Critical Entity Extractions

• $250 payment amount: Not extracted as FINANCIAL entity
• Upwork.com: Not identified as ORGANIZATION
• State of Georgia/Fulton County: Missing jurisdiction entities
• "Work for Hire": Critical legal concept not identified as LEGAL_CLAUSE

4. Semantic Analysis Contradictions

• Flagged "Ambiguous intellectual property ownership" as critical risk
• Document contains extensive, clear IP clauses including work for hire, assignment, power of attorney
• Suggests semantic analysis isn't properly reading the extracted entities

5. Metadata Quality Issues

Impact Assessment

These parsing failures undermine:

• Document classification reliability
• Entity extraction accuracy
• Legal analysis quality
• User trust in AI-powered document review

Expected Behavior

LexC should correctly identify:

1. Document type as "contract" or "consulting_agreement"
2. Both parties (Jennifer Liotta and Micha Schrom) as PERSON/PARTY entities
3. Financial terms ($250), organizations (Upwork.com), jurisdictions
4. Legal concepts ("Work for Hire", IP assignment clauses)
5. Populate metadata fields with extracted information

Reproduction

1. Process any standard consulting/service agreement
2. Examine LexC output for document_type, parties array, and entity extractions
3. Compare against document content for obvious misses

Environment

• Document format: .docx
• Content: Standard legal consulting agreement
• LexC version: Current production version
• Processing: Standard document intelligence pipeline

Related Issues

This may be related to the broader parsing issues mentioned in:

• #4735 (Data Parsing Failure Due to Incorrect Assumptions)
• #5428 (Terminal parsing failures due to ANSI contamination)

Suggested Investigation

1. Review document type classification logic
2. Audit party/person entity extraction rules
3. Examine confidence scoring algorithms
4. Test against broader corpus of standard legal documents
5. Validate semantic analysis integration with entity extraction

Priority: High - Core document processing functionality affected