recommendations

Recommendations Dataset

Source: synthetic_opus_demodbs (Neo4j text2cypher)

Total queries in original dataset: 753

Current Status

• queries.csv: 741 successfully converted and validated queries
• 12 queries that cannot be converted (documented below in Failed Queries)

Total: 741 + 12 = 753 ✓

Failed Queries (12 total)

Query 404

Reason: No weekday date component extraction in TypeQL.

MATCH (m:Movie)
WHERE date(m.released).weekday = 5
RETURN m.title, m.released
ORDER BY m.released DESC
LIMIT 3

Query 447

Reason: No epoch timestamp conversion (epochSeconds) in TypeQL.

MATCH (u:User)-[r:RATED]->(m:Movie)
WHERE m.released = date(datetime({epochSeconds: r.timestamp}))
RETURN u, r, m

Query 484

Reason: collect() aggregation has no direct TypeQL equivalent.

MATCH (m:Movie)<-[:DIRECTED]-(d:Director)
WITH d, m
ORDER BY m.imdbRating DESC
WITH d, collect(m)[0..5] AS topMovies
RETURN d.name AS director,
       [movie IN topMovies | movie.title] AS movies,
       [movie IN topMovies | movie.imdbRating] AS ratings

Query 537

Reason: No epoch timestamp conversion (epochSeconds) in TypeQL.

MATCH (u:User)-[r:RATED]->(m:Movie)
WHERE m.released = date(datetime({epochSeconds: r.timestamp}))
RETURN u.name AS user, m.title AS movie, m.released AS releaseDate
ORDER BY r.timestamp
LIMIT 3

Query 570

Reason: Duration functions (duration.between()) not available in TypeQL.

MATCH (m:Movie)<-[r:RATED]-(u:User)
WITH m, r, u,
     datetime(m.released) AS releasedDate,
     datetime({epochSeconds: r.timestamp}) AS ratingDate
WITH m, duration.between(releasedDate, ratingDate).days AS daysBetween
ORDER BY daysBetween
LIMIT 5
RETURN m.title AS movie, daysBetween AS daysBetweenReleaseAndRating

Query 577

Reason: No size(split()) — cannot count words in a string.

MATCH (m:Movie)<-[:DIRECTED]-(d:Director)
WHERE size(split(d.bio, ' ')) > 500
RETURN m.title
LIMIT 3

Query 591

Reason: No .year date component extraction — cannot compare release year to rating year.

MATCH (m:Movie)<-[r:RATED]-(u:User)
WHERE datetime(m.released).year = datetime(r.timestamp).year
WITH m, count(r) AS ratingCount
ORDER BY ratingCount DESC
LIMIT 5
RETURN m.title AS movie, ratingCount AS ratingsInFirstYear

Query 593

Reason: No size(split()) — cannot count quote delimiters in a string.

MATCH (m:Movie)
WHERE m.plot IS NOT NULL
RETURN m.title, m.plot
ORDER BY size(split(m.plot, '"')) DESC
LIMIT 5

Query 615

Reason: No date comparison functions — cannot compare release date to actor birth date.

MATCH (actor:Actor)-[:ACTED_IN]->(movie:Movie)
WHERE date(movie.released) = date(actor.born)
RETURN actor.name AS actor, movie.title AS movie, movie.released AS releaseDate

Query 626

Reason: No .year property accessor — cannot extract year from born date.

MATCH (d:Director)-[:DIRECTED]->(m:Movie)
WHERE m.year = d.born.year
RETURN d.name AS director, m.title AS movie, m.year AS year

Query 681

Reason: No size(split()) — cannot count words in a string.

MATCH (m:Movie)
WHERE size(split(m.plot, ' ')) > 100
RETURN m.title, m.plot
ORDER BY size(split(m.plot, ' ')) DESC
LIMIT 5

Query 738

Reason: Duration functions (date() - duration('P5Y')) not available in TypeQL.

MATCH (m:Movie)
WHERE m.released IS NOT NULL AND date(m.released) > date() - duration('P5Y')
RETURN m.title, m.released, m.imdbRating
ORDER BY m.imdbRating DESC
LIMIT 5

Original Cypher Errors

During conversion, TypeDB's richer type system and semantic review caught cases where the original Cypher did not correctly answer the English question.

Wrong counting target (1 query)

Index	Question	Cypher error
28	"First 5 movies rated by users from more than 10 different countries"	count(DISTINCT u.name) AS numCountries — counts distinct user names, not distinct countries. Variable is misleadingly named numCountries.

How TypeDB caught it: The TypeQL schema has an explicit located_in relation connecting users to countries. The TypeQL correctly counts distinct countries: located_in (resident: $u, country: $c); select $c; distinct; return count. Counting a generic property like u.name and calling it "countries" is not possible in a typed system.

Hardcoded workaround instead of schema attribute (1 query)

Index	Question	Cypher error
624	"Movies rated by both male and female users"	WHERE exists{ (m)<-[:RATED]-(:User {name: 'Omar Huffman'}) } AND exists{ (m)<-[:RATED]-(:User {name: 'Myrtle Potter'}) } — hardcodes two specific users instead of filtering by gender.

How TypeDB caught it: The TypeQL schema has an explicit gender attribute on User. The TypeQL correctly filters by gender: $u1 isa user, has gender "male"; $u2 isa user, has gender "female". The Cypher workaround assumes specific users represent genders, which fails for any other data.

Non-existent attribute reference (2 queries)

Index	Question	Cypher error
22	"Top 3 movies rated by the oldest users"	ORDER BY u.born ASC — User entity has no born attribute in the original schema (only Person does)
242	"First 3 movies rated by the youngest users"	ORDER BY u.born DESC — same non-existent attribute

How TypeDB caught it: TypeDB type-checks every attribute access against the schema. User cannot have born unless explicitly declared. The TypeQL schema added an age attribute to User, and the queries use sort $a desc (oldest = highest age) and sort $a asc (youngest = lowest age) respectively.