Supported Document Formats

The HyperText Markup Language (HTML) is the lingua franca of the web. Tika uses the jsoup library to support virtually any kind of HTML found on the web. The output from JSoupParser is guaranteed to be well-formed XHTML, and various heuristics keep things like inline scripts out of the extracted text.

The Extensible Markup Language (XML) is a generic container for all kinds of content. Tika has custom parsers for some widely used XML vocabularies like XHTML, OOXML and ODF, but the default DcXMLParser simply extracts the text content and ignores the XML structure. The one exception is Dublin Core metadata elements, which become document metadata.

Microsoft Office and some related applications produce documents in the generic OLE 2 Compound Document and Office Open XML (OOXML) formats. OLE 2 was the default from Office 97 until Office 2007 introduced OOXML. OfficeParser and OOXMLParser use the Apache POI libraries to extract text and metadata from both.

Old, pre-OLE2 Excel files (Excel 2, 3 and 4) are handled by OldExcelParser.

The older pure-XML office formats are handled by SpreadsheetMLParser, WordMLParser and Word2006MLParser.

OneNoteParser extracts text from OneNote notebooks, and MSOwnerFileParser pulls basic metadata from the temporary lock ("owner") files Office leaves beside open documents.

The OpenDocument format (ODF) is the native format of LibreOffice and OpenOffice. OpenDocumentParser supports the packaged format and the earlier OpenOffice 1.0 format on which ODF is based; FlatOpenDocumentParser handles the single-file flat XML variants (.fodt, .fods, .fodp).

The Apple iWork formats (Pages, Numbers, Keynote) are supported by IWorkPackageParser for the original package format. IWork13PackageParser and IWork18PackageParser identify the 2013 and 2018 generations and extract metadata, embedded resources and the preview image; the body text of these newer formats is not yet extracted.

The Corel WordPerfect Office formats are supported by WordPerfectParser (WordPerfect WP6 and later) and QuattroProParser (QuattroPro QPW v9 and later).

PDFParser parses PDF documents using the Apache PDFBox library. It can extract inline images, annotations, forms and incremental-update history, and can hand pages or images to OCR or a vision model. See PDF Parser for the options.

EpubParser supports the EPUB format used for many digital books. FictionBookParser supports the XML-based FictionBook (.fb2) publishing format.

RTFParser extracts text, metadata and embedded objects from Rich Text Format documents.

Tika uses the Commons Compress library to support most compression and packaging formats. CompressorParser handles the top-level compression formats (Gzip, BZip2, XZ, LZMA, Z, Brotli, LZ4, Snappy, …​), then PackageParser and its subclasses ZipParser and SevenZParser unpack the archive formats (Zip, 7Zip, Tar, AR, ARJ, CPIO, Dump, jar, …​) and pass each entry to a second parsing stage. See ZIP Detection and Salvaging for how damaged zips are handled.

RarParser supports the RAR archive format, which Commons Compress does not.

AppleSingleFileParser supports resources packaged in AppleSingle and AppleDouble files, and PListParser reads Apple property lists in both XML and binary form.

Extracting text from plain-text files looks simple until character encodings enter the picture. TextAndCSVParser runs Tika’s encoding detectors (see Encoding Detectors) and recognizes CSV and TSV, emitting tables for those.

MarkdownParser parses Markdown into structured XHTML rather than treating it as plain text.

FeedParser supports the RSS and Atom feed formats. IptcAnpaParser supports the IPTC ANPA news wire format.

WARCParser and WACZParser unpack WARC and WACZ web archives and parse each captured response; HttpParser handles a single raw HTTP response.

ChmParser supports the CHM (compiled HTML help) format.

Tika extracts metadata from common audio formats, and text from those that carry lyrics or other textual content. Extracted metadata includes sampling rates, channels, format information, artists, titles and so on. AudioParser and MidiParser use the standard javax.sound features for simple formats. Mp3Parser adds MP3, and MP4Parser covers MP4 audio. The Ogg family (Vorbis, Speex, Opus, FLAC) is supported by VorbisParser, OpusParser, SpeexParser and FlacParser.

Speech-to-text is available through AmazonTranscribe in the tika-transcribe-aws module of tika-parsers-ml; it is not part of the standard set.

ImageParser uses the standard javax.imageio feature to extract simple metadata from the image formats the Java platform supports, such as PNG, GIF and BMP. Richer metadata comes from JpegParser and TiffParser, which use the metadata-extractor library for Exif, IPTC and XMP. HeifParser (HEIF/HEIC), JXLParser (JPEG XL), WebPParser, PSDParser, BPGParser and ICNSParser extract metadata from their respective formats.

WMFParser and EMFParser extract text from Microsoft WMF and EMF drawings; EMFParser also exposes embedded resources.

Text inside images is not extracted by these parsers. See OCR and vision-language models.

FLVParser supports the Flash video format with a simple parsing algorithm. MP4Parser covers the MP4 family (MP4, QuickTime, 3GPP, …​), extracting video metadata and any audio stream metadata. For the Ogg family, OggParser and TheoraParser extract a limited amount of metadata.

Frame extraction, transcoding and probing through ffmpeg are available via the External Parser when the tool is installed.

ClassParser extracts class names and method signatures from Java class files, and PackageParser supports jar archives.

SourceCodeParser handles a number of source code formats, including Java, C, C++ and Groovy, providing a formatted form of the code along with simple metadata.

MboxParser extracts messages from the mbox format used by many mail archives and Unix mailboxes. RFC822Parser processes single RFC 822 messages (.eml). OutlookPSTParser extracts messages from Microsoft Outlook PST stores, and OfficeParser handles single Outlook MSG files. TNEFParser extracts attachments from Microsoft TNEF (winmail.dat) wrappers.

DWGParser extracts metadata from AutoCAD DWG files, DGN8Parser from MicroStation DGN v8 files, and PRTParser from Cadkey PRT files.

TrueTypeParser extracts simple metadata from TrueType fonts, and AdobeFontMetricParser does the same for Adobe Font Metrics files.

IDMLParser reads Adobe InDesign Markup (IDML) packages, and MIFParser reads Adobe FrameMaker Interchange Format. HwpV5Parser supports the Hangul Word Processor (HWP) v5 format.

For translation memory and localization interchange, XLIFF12Parser and XLZParser handle XLIFF 1.2 and its zipped form, and TMXParser handles Translation Memory eXchange.

In the standard set, DIFParser extracts attribute metadata from the GCMD Directory Interchange Format and MatParser from MATLAB .mat files.

The tika-parser-scientific-package artifact in tika-parsers-extended adds GDALParser, GeographicInformationParser (ISO 19139), GribParser, HDFParser, ISArchiveParser (ISA-Tab) and NetCDFParser. These pull in large native and scientific dependencies and are not in tika-app.

ExecutableParser extracts platform, architecture and type metadata from a range of executable and library formats, such as Windows PE and Linux/BSD ELF binaries; UniversalExecutableParser unpacks macOS universal (fat) binaries.

Pkcs7Parser parses the content of PKCS7 signed messages, without information from the outer PKCS7 wrapper. See Detectors for enabling finer CMS detection. TSDParser extracts metadata from Time Stamped Data envelopes and exposes the wrapped content.

JackcessParser extracts metadata and tabular content from Microsoft Access databases. DBFParser supports dBase (.dbf) files before version 7, as used by dBase, FoxBASE, FoxPro and ESRI Shapefiles. SAS7BDATParser reads SAS .sas7bdat datasets.

SQLite3Parser extracts SQLite3 content in tabular form. It ships in the tika-parser-sqlite3-package artifact in tika-parsers-extended because of the native sqlite-jdbc dependency; it is not in tika-app.

None of the image parsers above read text out of pixels. Tika offers three ways to do that, all off by default and all documented under Configuration:

The PDF Parser page describes how these engines are attached to PDF pages and images.

The tika-parser-nlp-package artifact in tika-parsers-ml wraps external NLP tools as parsers that enrich the output of other parsers: NamedEntityParser (named entity recognition via OpenNLP, Stanford CoreNLP, MITIE, NLTK or regex), JournalParser (journal article metadata via a Grobid server), CTAKESParser (clinical text via Apache cTAKES) and GeoParser (geographic entity resolution against a gazetteer). Each needs models, training data or a running service; see the parser javadocs.

Generated from the parsers on the tika-app classpath with tika-app --list-parser-details-adoc. Parsers that only advertise types once an external tool or service is configured (Tesseract, the VLM parsers, the External Parser) are not listed.