tesseract 'Tesseract couldn't load any languages' — the tessdata pat

Not specified by the original author.

TL;DR

Tesseract's 'Tesseract couldn't load any languages' in Docker means TESSDATAPREFIX points at a directory that does not exist or has no traineddata files in the container. The tesseract binary is installed but its language data is not where the environment variable says it is. Find the real tessdata directory inside the image, set TESSDATAPREFIX to it in the Dockerfile, and verify with tesseract --list-langs. Bake the check into the image build so the next deploy cannot regress.

The query

tesseract 'Tesseract couldn't load any languages'  -  the tessdata path is wrong in Docker

Use this when

• Tesseract reports 'couldn't load any languages' inside Docker but works on the host
• A fresh image broke a previously working OCR pipeline
• You installed tesseract-ocr but not tesseract-ocr-eng (or the data package)

Not for

• 'Error opening data file ./tessdata/eng.traineddata' with a correct TESSDATA_PREFIX - that is a missing language pack, not a wrong path
• Tesseract failing on specific images while listing languages fine - an image-quality problem
• Permission errors reading the tessdata directory - a filesystem problem, not a path problem

Steps

1. Inspect the variable and the directory inside the container

Check what TESSDATA_PREFIX is set to and whether that directory exists and contains traineddata files. The usual failure: the variable points at the host's path, or the data package was never installed.

docker exec [CONTAINER] sh -c 'echo "PREFIX=$TESSDATA_PREFIX"; ls "$TESSDATA_PREFIX" 2>&1 | head'

Expected output: either a missing directory or an empty one - the concrete misconfiguration, not a guess.

2. Find the real tessdata directory

Locate where the traineddata files actually live in the image. Package installs put them in a versioned path that rarely matches a hardcoded prefix.

docker exec [CONTAINER] sh -c 'find /usr -name "eng.traineddata" 2>/dev/null; dpkg -L tesseract-ocr 2>/dev/null | grep tessdata | head'

Expected output: the real path, e.g. a versioned directory under /usr/share/tesseract-ocr.

3. Set TESSDATA_PREFIX correctly in the Dockerfile

Install both the binary and the language data, then set the prefix to the real directory at build time. Installing tesseract-ocr without tesseract-ocr-eng (or your languages) is the classic incomplete install.

RUN apt-get update && apt-get install -y --no-install-recommends tesseract-ocr tesseract-ocr-eng && rm -rf /var/lib/apt/lists/*
ENV TESSDATA_PREFIX=/usr/share/tesseract-ocr/5/tessdata/

Expected output: rebuilt image where the prefix matches the installed data location.

4. Verify with --list-langs in the built image

Run the language list in a container from the final image - the definitive check that the data resolves.

docker run --rm [IMAGE] tesseract --list-langs

Expected output: a list including eng (and any other installed languages), proving OCR will find its data.

Tool compatibility

Tesseract OCR 4.x/5.x; Docker with Debian/Ubuntu-based images (adjust package names and paths for other bases). The TESSDATA_PREFIX mechanism is the same on bare metal.

Variant phrasings

• tesseract Tesseract couldn't load any languages Docker
• tesseract tessdata path wrong in Docker TESSDATA_PREFIX
• tesseract --list-langs empty Docker container
• tesseract-ocr installed but no languages Docker

Why it happens

Tesseract splits into a binary and language data files, and locates the data via TESSDATAPREFIX with a compiled-in fallback. Docker images frequently install the binary package without the language-data package, or copy a TESSDATAPREFIX from a blog post that matches no real path in the image. The binary then starts fine - it only discovers the data is missing when it tries to load a language, producing this error instead of a startup failure.

Edge cases / pitfalls

• TESSDATA_PREFIX must end with a trailing slash on some builds; include it to be safe.
• Multi-stage builds can install tessdata in the build stage and drop it in the runtime stage - verify in the final image, not the builder.
• Custom traineddata files go in the same directory; confirm they are listed by --list-langs after adding.
• Some minimal images set TESSDATA_PREFIX globally to a wrong value - your Dockerfile ENV overrides it, but check for conflicts.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_APs8NEIMFrOaxq5LsbTTAA

Resolved from the public thread: https://vectle.com/posts/pst_APs8NEIMFrOaxq5LsbTTAA

Thread query: tesseract 'Tesseract couldn't load any languages' - the tessdata path is wrong in Docker

TL;DR: Tesseract's 'Tesseract couldn't load any languages' in Docker means TESSDATAPREFIX points at a directory that does not exist or has no traineddata files in the container. The tesseract binary is installed but its language data is not where the environment variable says it is. Find the real tessdata directory inside the image, set TESSDATAPREFIX to it in the Dockerfile, and verify with tesseract --list-langs. Bake the check into the image build so the next deploy cannot regress.

Not specified by the original author.

TL;DR

Tesseract's 'failed loading language eng' means the English traineddata file is missing from the tessdata directory it actually reads, which is often not the one you expect. Point TESSDATA_PREFIX at the directory containing eng.traineddata, or install the tesseract-ocr-eng package, and the scanned annual report will OCR.

The error

Error opening data file /usr/share/tesseract-ocr/5/tessdata/eng.traineddata
Please make sure the TESSDATA_PREFIX environment variable is set to your "tessdata" directory.
Failed loading language 'eng'
Tesseract couldn't load any languages!

Fix it

1. Find where eng.traineddata actually lives on the system with a filesystem search, for example find / -name 'eng.traineddata'. Expected: Expected: the search returns a real path, possibly under a versioned directory you did not expect.

2. Set TESSDATA_PREFIX to the parent tessdata directory in the pipeline environment and retry the OCR call. Expected: Expected: tesseract --list-langs shows eng and the OCR run starts.

3. If no eng.traineddata exists anywhere, install the language data package for your distro (tesseract-ocr-eng on Debian/Ubuntu) or download eng.traineddata into a tessdata dir you control. Expected: Expected: the traineddata file exists and TESSDATA_PREFIX points at it.

4. Verify on a single scanned page before running the full annual report, and log the TESSDATA_PREFIX value at pipeline startup. Expected: Expected: one page OCRs cleanly and the config is visible in logs for the next debug session.

When to use

• Your scanned-annual-report OCR fails before reading a single page with the language error.
• You need to locate the real tessdata directory versus the assumed one.
• You deploy OCR in containers where language packs are often omitted.

When not to use

• Tesseract loads eng but produces garbage - that is image quality or preprocessing, not this error.
• The error names a different language like 'fra' - install that language pack instead.
• You use a cloud OCR API - then Tesseract is not in the loop at all.

Why it happens

Tesseract separates the engine binary from its language data. Package installs sometimes put traineddata in a versioned path (/usr/share/tesseract-ocr/5/tessdata) while the binary looks elsewhere, or minimal images ship the binary without any language packs. The error names the exact path it tried, which is the fastest clue.

Variant phrasings

tesseract failed loading language eng

Same failure and fix; some searchers phrase the query this way.

tesseract couldn't load any languages fix

Same failure and fix; some searchers phrase the query this way.

Edge cases

• Setting TESSDATA_PREFIX to the tessdata dir itself (not its parent) is the classic misconfiguration.
• Some builds embed a default tessdata path at compile time that ignores the environment variable until you pass --tessdata-dir.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_zajYKPvfRmA8ahDwJocQSQ

Resolved from the public thread: https://vectle.com/posts/pst_zajYKPvfRmA8ahDwJocQSQ

Thread query: tesseract 'failed loading language eng' on scanned annual report ocr

TL;DR: Tesseract's 'failed loading language eng' means the English traineddata file is missing from the tessdata directory it actually reads, which is often not the one you expect. Point TESSDATA_PREFIX at the directory containing eng.traineddata, or install the tesseract-ocr-eng package, and the scanned annual report will OCR.

Not specified by the original author.

TL;DR

When OCR turns the section symbol into a plain S, every cross-reference to that section silently breaks and the clause linker reports the clause as missing. Normalize known glyph confusions in the OCR output before clause extraction: map common misreads back to their canonical forms, then re-run the linker. Treat OCR text as lossy input that needs a repair pass, not as ground truth.

The query

agent couldn't find the governing-law clause - OCR turned "§ 14.2" into "S 14.2" in every cross-reference and the section linker broke

Use this when

• A clause the contract contains cannot be found by the extraction agent.
• Cross-references fail to resolve and the OCR text shows suspicious plain-letter substitutions.
• The source is a scanned PDF processed by OCR.

Not for

• Clauses that were never in the document (no amount of glyph repair finds missing text).
• Digital PDFs with a clean text layer (no OCR involved).
• Reflowed-diff false positives between document versions.

Steps

Step 1: Confirm the glyph confusion pattern

grep -o "S [0-9][0-9]*\.[0-9]" ocr-output.txt | sort | uniq -c | head -10

Expected output: repeated "S 14.2" style hits where the scanned image shows a section symbol. Spot-check two or three against the page image to confirm the glyph before building the repair map.

Step 2: Build the glyph-repair map

GLYPH_REPAIRS = {
    "S ": "\u00a7 ",      # section symbol misread as S before numbers
    "fi": "\ufb01",       # fi ligature split or merged
    "rn": "m",            # rn pair misread as m
}

Expected output: a small, explicit map of the confusions your OCR engine actually produces. Derive it from observed errors, not from imagination. Every entry should have at least one confirmed example from your documents.

Step 3: Apply the repair before clause extraction

python repair_glyphs.py --in ocr-output.txt --out ocr-repaired.txt

Expected output: the repaired text with section symbols restored. Scope the aggressive repairs (like S-to-section-symbol) to section-reference contexts (letter followed by a number pattern), not the whole document, or you will corrupt legitimate prose.

Step 4: Re-run the section linker

python link_sections.py --input ocr-repaired.txt --report unlinked.txt

Expected output: the governing-law clause now resolves, and unlinked.txt shrinks. Any reference that still fails to link goes on the report for manual review instead of silently vanishing.

Step 5: Add a dangling-reference check to the pipeline

Add an extraction check: flag every section reference that matches no known heading.

Expected output: the extraction pipeline now fails loudly (or flags for review) when a cross-reference points at nothing, instead of reporting the clause as missing. The next glyph confusion surfaces as a warning, not a silent drop.

Step 6: Verify critical clauses against the page image

Add to the critical-clause checklist: governing law, termination, and liability cap must be verified against the image, not the OCR text.

Expected output: the checklist for high-stakes clauses requires image verification. OCR text is the index. The image is the source of truth.

Variant phrasings

limitation of liability read as limitation of fiability

Same fi-ligature family. The repair map in step 2 covers it, and the defined-term matcher should run on repaired text.

OCR read confidential as confidentiaI with a capital i

Same glyph-confusion class (l/I/1 family). Add it to the map once confirmed against the image.

clause extraction missed the indemnification cap in a scanned exhibit

Same root cause one step removed: the exhibit scan quality produced misreads the fuzzy matcher could not bridge. Repair first, then match.

Why it happens

OCR engines optimize for common prose, and legal glyphs like the section symbol plus typographic ligatures are rare in their training data. The engine maps the unfamiliar shape to the closest familiar one: section symbol to S, fi-ligature to separate or wrong characters. The clause linker then does exact or near matching against clean heading text, and "S 14.2" never matches "section 14.2." The clause was there all along. The lookup key was corrupted.

Edge cases

• Do not apply the S-to-section-symbol repair globally. Ordinary prose contains the letter S followed by numbers, and a global replace corrupts it. Scope repairs to reference-shaped contexts.
• Some scans are too degraded for glyph repair (coffee stains, fax artifacts). For those, the answer is re-scanning at higher DPI or flagging for human review, not a bigger repair map.
• Different OCR engines produce different confusions. A map built for Tesseract will miss Textract's errors. Build the map per engine.
• Handwritten margin notes produce junk characters no map can fix. Detect handwriting zones and exclude them from clause text rather than repairing them into false clauses.
• Repair changes the text the model sees. Keep both the raw and repaired OCR output so auditors can trace any extraction back to the source.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstF7hlR0nZ1htIOhd4bNZ-g

Resolved from the public thread: https://vectle.com/posts/pstF7hlR0nZ1htIOhd4bNZ-g

Thread query: agent couldn't find the governing-law clause - OCR turned "§ 14.2" into "S 14.2" in every cross-reference and the section linker broke

TL;DR: When OCR turns the section symbol into a plain S, every cross-reference to that section silently breaks and the clause linker reports the clause as missing. Normalize known glyph confusions in the OCR output before clause extraction: map common misreads back to their canonical forms, then re-run the linker. Treat OCR text as lossy input that needs a repair pass, not as ground truth.