Maps to GROUP BY + aggregate + subquery-based FILTER on MAX/MIN to handle ties. Computes an aggregate across ≥2 groups and returns the entity (or entities) with the extreme aggregate value (e.g. 'Which department has the highest average salary?'). Comparative question Example: 'in Berlin departments managers list'. Example: 'salary employees highest department which'. Example: 'employees high sales?'. Ungrammatical or non-canonical syntactic structures affecting parsing and dependency analysis. Syntactic distortion A list question where the answer is a sample of an unordered enumeration of entities or facts about entities. Maps to SELECT...SAMPLE... Sampled list question Answer is a single numeric value resulting from an aggregation (MAX, MIN, AVG, COUNT, SUM). Scalar answer Example: 'John' → multiple people named John. Example: 'Paris' → city vs mythology entity. Ambiguity where a surface form maps to multiple candidate RDF entities. Entity linking ambiguity Relates a question type to the class of answer it expects. has answer type A list question where each item pairs a group with its computed aggregate value (e.g. 'What is the average salary per department?'). Maps to SELECT ?group (AGG(?x) AS ?val) … GROUP BY ?group. Aggregation list question A list question whose answer is ordered by an existing property, usually with an explicit cardinality limit (e.g. 'Who are the top 3 employees by seniority?'). Maps to SELECT … ORDER BY DESC(?prop) LIMIT n. No aggregation required. Limited ranked list question Answer is YES/TRUE or NO/FALSE. Boolean answer A question expecting a boolean YES/NO answer. Maps to ASK { … } with BGPs. Confirmation question Example: 'averge revenu of compnies'. Example: 'emplyees with hgh salery'. Spelling mistakes, typos, ASR errors, or MT artifacts affecting token-level interpretation. Orthographic noise Expects a deduplicated set of entities or facts (e.g. 'Which countries export cocoa?'). Maps to SELECT DISTINCT for plain enumerations, or GROUP BY for aggregate enumerations (qat:AggregateList). Both guarantee a set without duplicate rows. Enumeration question Answer is a single named entity or literal fact. Single entity answer Answer is an ordered list of entities sorted by an existing property, bounded by a cardinality limit. Ranked list answer Answer is the tuple describing the entity that wins a comparative aggregation across groups. Specialises both ScalarAnswer (results from aggregation) and EntityListAnswer (identifies entities). Ranked entity answer Example: 'manager nationality' not modeled in schema. Example: 'employee satisfaction score' not in schema. Required concept is absent or not modeled in the ontology or knowledge base. Schema coverage mismatch Answer is an enumeration of (group, aggregate value) pairs. Specialises both ScalarAnswer (each pair contains an aggregate) and EntityListAnswer (groups are entities). Aggregate list answer A question whose answer requires an aggregation operation (MAX, MIN, AVG, COUNT, SUM) over a set of values. Maps to SELECT (AGG(?x) AS ?val) WHERE. Aggregate factoid question Example: 'Which companies acquired startups and where are they located?'. Example: 'Who is John's manager and what is his salary?'. Failures in resolving pronouns, ellipsis, or implicit discourse references. Discourse reference failure Example: 'number of employees in each department'. Example: 'top selling products'. Ambiguity in aggregation operators, grouping, ranking, or aggregation scope (COUNT, SUM, ORDER BY). Aggregation underspecification Example: 'avg temp in US states'. Example: 'GDP growth in EU'. Underspecified abbreviations or acronyms requiring expansion or contextual disambiguation. Abbreviation ambiguity A list question whose answer is ordered by an existing property (e.g. 'Who are the employees by seniority?'). Maps to SELECT … ORDER BY DESC(?prop). No aggregation required. Ranked list question A question expecting as answer one or more facts. Typically introduced by a wh-word. Maps to a plain SELECT … WHERE pattern. Factoid question Answer type Abstract superclass for all question type categories. Question type Example: 'products with volume < 15 cm3' → computed from length × width × height. Example: 'company size' → number of employees. Example: 'customer age' → derived from birthDate. Concept is present but requires transformation or derived computation from available attributes. Entity/attribute mismatch Example: 'recent publications'. Example: 'high salary'. Example: 'young employees'. Predicates lacking explicit thresholds or formal schema grounding requiring external interpretation. Vague predicate grounding This layer complements lsqv:hasStructuralFeatures analysis by capturing mismatches between intended meaning, RDF modeling, and query structure. Issues related to meaning representation, schema alignment, ontology mapping, and SPARQL structural-semantic correspondence. Semantic issue Example: 'number of employees in sales or marketing who are managers' → different AND/OR groupings. Example: 'average salary of employees in hardware departments' → per-department vs global average. Example: 'departments where employees earn more than 50k' → all vs at least one employee. Ambiguity in quantification, scope, and logical composition (AND/OR/NOT/OPTIONAL). Logical form ambiguity Example: 'companies that make cars' → automotive manufacturers. Example: 'pay' → salary. Example: 'French poets' → poets with France as nationality/affiliation. Lexical variation requiring normalization such as synonyms, paraphrases, or compositional expressions. Lexical grounding mismatch Answer is an unordered enumeration of entities or facts. Entity list answer This layer captures failures in surface form interpretation before or alongside schema grounding. Issues related to linguistic realization of the question, including syntax, lexical choice, ambiguity, and phrasing. Linguistic issue A question that asks about entities or facts that lack a certain property or relationship, using FILTER NOT EXISTS to exclude results where a given pattern holds. Counterfactual question Answer is a sample of an unordered enumeration of entities or facts. Sampled list answer QA Question Types 2026-04-08 QA Question Types taxonomy Taxonomy of natural language question types for QA systems, extending Qanary. Question types are defined based on expected answer shape, structural features, and LSQ feature alignment. SPARQL structural signatures are expressed via lsqv:usesFeature. Ontology structure and taxonomy co-edited interactively with Claude Sonnet 4.6 (Anthropic) on 2026-04-08. Question type hierarchy, disjointness axioms, and vocabulary alignment with lsqv: and sp: developed interactively. The ontology also defines a lightweight diagnostic issue layer for question-level QA error analysis, aligned with Qanary annotations and Open Annotation targeting/body relations. 1.0 Example: 'People married in 2020' → requires join through marriage event and date. Example: 'employees with skills in data science' → employee → hasSkill → skillCategory. Example: 'managers in departments in Berlin' → manager → department → location. Missing explicit predicates requiring multi-hop traversal or inferred joins between entities. Implicit relation inference