The timeout option was ignored when using the chemaxon.naming.document.TextToStructure API.

Processing was getting stuck in the presence of some names with a very large number of brackets.

The processing of images in the JBIG2 format in PDF documents for OCR and OSR has been temporarily disabled. It will be re-enabled by default once we require Java 8, which we expect to do in a few months time. In the mean time, if this feature is desired, a Java ImageIO JBIG2 plugin can be added to the classpath to enable this feature. Please contact us if you have any issue with this.

Some incorrect SMILES strings were detected.

Complex IUPAC names are now detected even inside a bracketed subsentence.

The new version of method TextToStructure.extract accepts a DocumentToStructureOptions instead of an option string, allowing higher-level usage and faster processing.

Processing of short text using chemaxon.naming.document.TextToStructure is about 10% faster.

Class chemaxon.naming.document.D2S.Options is renamed to chemaxon.naming.document.DocumentToStructureOptions.

When processing short XML fragments, passing the option content-type=text/xml now leads to higher performance by skipping format detection entirely.

Option content-type was ignored in chemaxon.naming.document.TextToStructure

The "Preparation N" identifier is now automatically detected in patents, in additions to exemplified compounds.

PDF documents OCRed by PDF-XChange are not OCRed a second time, avoiding duplicate results and resulting in much faster processing.

The 'end' character position property was missing or incorrect in names delimited by ":" characters.

Structures mentioned in plural, for instance "pyrimidines", are now tagged with the "generic" type.

The format of HTML documents can now be detected from the file contents only.

The -groups option did not work for common names.

The format of Microsoft Office documents can now be detected from the file contents only.

A trailing name between brackets, often a synonym of the previous name, for instance "retinoic acid (vitamin A)", was sometimes misrecognized as a part of the previous name.

Example numbers are now detected in Chinese language patents.

? (water) is now recognized as a vernacular term in Chinese and Japanese.

More syntaxes for example numbers are detected in patents.

A spurious warning was logged when recognizing an InChI string in a document.

Some Japanese names were not recognized without the Chinese Name to Structure license.

Some Asian names were wrongly starting with a closing bracket.

The detection of Asian names in documents is improved.

Some invalid structures returned by an OSR tool could interrupt document to structure. They are now simply logged and processing continues.

Asian names followed by some number formats were not recognized.

Some names with spurious whitespace are better detected in documents.

When detecting exemplified compound numbers in patents, more syntaxes are now detected.

When processing Office documents with Optical Structure Recognition enabled, the following log message was issued: WARNING: Skipping non-existing image.

The warning "JBIG2ReadParam not specified. Default will be used.", which appeared in the logs, has been fixed.

The detection of asian names is improved in some documents.

Processing of scanned PDFs (using OCR - Optical Character Recognition) was failing on Mac OS X.

Names that contain extra spaces (for instance because of OCR or new lines without a - character before the break) are now better supported.

Patent PDFs provided by LexisNexis's Univentio can now be processed using the built in OCR text instead of doing the OCR again, leading to about 10 times faster processing. This was already working for older Univentio PDFs but is now also supported for recent ones.

Improvements in OCR error correction.

Significant improvements are included in OCR error correction, especially for patents based on scanned images.

Exemplified compound numbers are now extracted automatically when they are mentioned directly before the IUPAC name.

Improved detection of asian names in documents.

Some TIFF images failed to be processed.

Certain "custom" image types in PDFs failed to be processed by OSR tools.

The location of OSR tools can now be specified by using Java system properties: chemaxon.naming.clide.path, chemaxon.naming.osra.path and chemaxon.naming.imago.path. These take precedence over the environment variables (CLIDE, OSRA and IMAGO), which are also supported.

The detection of Japanese names in documents has been improved.

In the output of Optical Structure Recognizers, some non-standard labels such as X2 are now interpreted as R-groups. Those structures also become representable as SMILES.

Some aliases detected by Optical Structure Recognition programs are expanded to the corresponding chemical group.

Encrypted documents are now reported with a clearer message and a specific single exception: chemaxon.naming.document.EncryptedDocumentException

Failure to start OCR could lead to multiple entries in the log.

A failure in writing temporary images to disk lead to duplicate entries in the log.

Unknown format options now lead to a failure instead of being ignored.

Numbers in subscript and superscript HTML and XML tags (<sub> and <sup>) are now interpreted.

Japanese and Chinese names were not recognized in documents when followed by a fullwidth number inside brackets.

Assignee metadata was not extracted from some recent USPTO XML patents.

Numeric character entities in metadata of XML documents are now decoded.

Some SMILES-like words containing "Br" or "Cl" were wrongly detected as SMILES.

The structures recognized by OSR tools (CLiDE and OSRA) are now represented in a uniform way, for instance using real R-group atoms instead of just aliases.

Some OLE structures embedded by JChem for Office were extracted as empty structures.

A failure to process some Powerpoint documents was fixed.

Logging messages are improved to facilitate investigating problematic documents.

Detection of Chinese chemical names in documents was improved.

Names (especially Chinese) are better detected in XML patent files from IFI Claims.

The molecule name extracted from Chinese and Japanese documents was in a few cases not the actual name but another close-by word.

Some invalid structures were generated from some chinese and japanese documents with unusual character combinations.

The embedded OCR information in PDFs from the TotalPatent system (by LexisNexis) is now detected and used automatically. This leads to massively faster processing for such PDFs, from 10 to 30 times faster!

Names split between two paragraphs can now be detected and converted to structure. While ideally this situation should not occur in proper semantic documents, it does occur in automatically generated ones, for instance in patent documents distributed by patent providers.

The context field can now include text from a previous paragraph. This is useful in particular in patents, where this sometimes includes the example number.

Google patent PDFs were only processed until page 20.

Processing an HTML file with CLiDE enabled failed when an image contained no structure.

Added the osrResolution=N option to specify the resolution in DPI of images processed by OSRA. This is only needed for single image files, not for images inside PDF, Office or HTML documents.

The character encoding of HTML5 documents was not detected automatically.

Chinese and Japanese name are better detected in documents.

Optical Structure Recognition is now performed on images referenced in HTML documents. This is supported when the image is accessible as a file (relatively to the HTML document file), and when the image is represented by a data URI.

The BMP image format is now supported for calling Optical Structure Recognition tools using the document to structure API.

The ending position of the names in text documents (HTML, XML, TXT) is now also included in the properties of returned structures.

Some embedded structures in Powerpoint documents were not extracted.

OSRA 2.0 is supported.

A format option has been introduced allowing the user to select an optical structure recognition tool: CLiDE, OSRA or Imago when more than one is installed on the computer. For instance, d2s:osra will request OSRA to be used.

The encoding of HTML files was not always detected from the META tag.

When the extraction of OLE embedded structures from Office documents was disabled using the d2s:-ole format option, the optical structure recognition of images was disabled as well.

The character position field of structures was sometimes higher by a few characters when extracted from HTML documents with CRLF (\r\n) line endings.

The CLiDE optical structure recognition tool was not automatically detected on 64 bit versions of Windows.

When using OSRA, some structures with implicit hydrogens were wrongly filtered out.

After processing only a part of an HTML document, the processing of the next HTML document could give some incorrect results.