Given a DOCX with Zotero citations, how can I convert its contents to LaTeX?

[tommyvct]

I wasted a whole day researching on this but there is literally less than 5 related results on the whole internet:

• https://retorque.re/zotero-better-bibtex/citing/migrating/index.html
• help converting docx with Zotero cites to latex #10904
• https://forums.zotero.org/discussion/108968/pandoc-citation-format-as-a-citation-style-with-betterbibtex

and maybe some stack exchange unanswered questions.

From the very, very, very poor written migration guide from DOCX to Markdown by Better BibTeX I tried:

pandoc -f docx+citations -t markdown -i "Thesis Proposal Real.docx" -o proposal.md

and

pandoc -f docx+citations -t latex -i "Thesis Proposal Real.docx" -o proposal.tex 

and this is what I got:

Nishida and Nakayama (2020)[@2018] adapt unsupervised syntactic parsing
(Viterbi EM) to discourse by hypothesizing that discourse and syntax
share similar constituent regularities. 

and for LaTeX

Nishida and Nakayama (2020){[}51{]} adapt unsupervised syntactic parsing
(Viterbi EM) to discourse by hypothesizing that discourse and syntax
share similar constituent regularities.

Notice how the markdown citation tag differs from the LaTeX one, and from the original DOCX rendered in Microsoft Word:

I also tried the CSL format and it only made it worse:

Nishida and Nakayama (2020){[}{[}CSL STYLE ERROR: reference with no
printed form.{]}{]} adapt unsupervised syntactic parsing (Viterbi EM) to
discourse by hypothesizing that discourse and syntax share similar
constituent regularities. 

To reply to the original comment by @iandol , when I tried to export the library to Quick Look, All I get is a file with a pair of []. Note that I have set the quick copy format to pandoc citation in the Better BibTeX settings.

To clarify, I want to convert DOCX with Zotero citations, to LaTeX. The LaTeX file should have citations readily available in the standard format of \cite{yang_xlnet_nodate}. Pandoc documentation claims that the citations plugin of DOCX can handle Zotero citations, but there is absolutely no meaningful documentation and examples whatsoever from them how this works.

I have the .bib exported from Zotero with good usable tags with it, like this:

@article{yang_xlnet_nodate,
	title = {{XLNet}: {Generalized} {Autoregressive} {Pretraining} for {Language} {Understanding}},
	abstract = {With the capability of modeling bidirectional contexts, denoising autoencoding based pretraining like BERT achieves better performance than pretraining approaches based on autoregressive language modeling. However, relying on corrupting the input with masks, BERT neglects dependency between the masked positions and suffers from a pretrain-finetune discrepancy. In light of these pros and cons, we propose XLNet, a generalized autoregressive pretraining method that (1) enables learning bidirectional contexts by maximizing the expected likelihood over all permutations of the factorization order and (2) overcomes the limitations of BERT thanks to its autoregressive formulation. Furthermore, XLNet integrates ideas from Transformer-XL, the state-of-the-art autoregressive model, into pretraining. Empirically, under comparable experiment setting, XLNet outperforms BERT on 20 tasks, often by a large margin, including question answering, natural language inference, sentiment analysis, and document ranking.1.},
	language = {en},
	author = {Yang, Zhilin and Dai, Zihang and Yang, Yiming and Carbonell, Jaime and Salakhutdinov, Russ R and Le, Quoc V},
	file = {PDF:/Users/tommyvct/Zotero/storage/E3TSZ7YX/Yang et al. - XLNet Generalized Autoregressive Pretraining for Language Understanding.pdf:application/pdf},
}

while if I use Better BibTeX, this is what I get:

@article{
  title = {{{DiMLex}}: {{A Lexicon}} of {{Discourse Markers}} for {{Text Generation}} and {{Understanding}}},
  author = {Stede, Manfred and Umbach, Carla},
  abstract = {Discourse markers ('cue words') are lexical items that signal the kind of coherence relation holding between adjacent text spans; for example, because, since, and for this reason are different markers for causal relations. Discourse markers are a syntactically quite heterogeneous group of words, many of which are traditionally treated as function words belonging to the realm of grammar rather than to the lexicon. But for a single discourse relation there is often a set of similar markers, allowing for a range of paraphrases for expressing the relation. To capture the similarities and differences between these, and to represent them adequately, we are developing DiMLex, a lexicon of discourse markers. After describing our methodology and the kind of information to be represented in DiMLex, we briefly discuss its potential applications in both text generation and understanding.},
  langid = {english},
  file = {/Users/tommyvct/Zotero/storage/NB5A5FNW/Stede and Umbach - DiMLex A Lexicon of Discourse Markers for Text Generation and Understanding.pdf;/Users/tommyvct/Zotero/storage/NPN3NEW8/Stede and Umbach - DiMLex A lexicon of discourse markers for text generation and understanding.pdf}
}

Please help.

Thanks.

[jgm]

Why don't you upload the docx, or a reduced version of it that suffices for an example.

[tommyvct]

Thesis Proposal Real.docx
here it is

cc @iandol

I created this document well before I installed Better BibTeX.

[iandol]

This docx indeed doesn't have any cite-keys in it, thus only ID is available...

{"citationID":"8R746EE9","properties":{"formattedCitation":"[2]","plainCitation":"[2]","noteIndex":0},"citationItems":[{"id":2495,"uris":["http://zotero.org/users/17530164/items/ILAMPVSR"],"itemData":{"id":2495,"type":"thesis","publisher":"Université de Lorraine","title":"Facing Data Scarcity in Dialogues for  Discourse Structure Discovery and  Prediction","author":[{"family":"Li","given":"Chuyuan"}],"issued":{"date-parts":[["2023"]]}}}],"schema":"https://github.com/citation-style-language/schema/raw/master/csl-citation.json"}

[iandol]

One problem is the following:

1. Zotero citation inserted into word with BetterBibTeX installed:

"id":13159 and "citation-key":"lamme2018" are both present in the Zotero injected JSON:

{ADDIN ZOTERO_ITEM CSL_CITATION {"citationID":"9zoD6v0b","properties":
{"unsorted":false,
"formattedCitation":"(Lamme, 2018)",
"plainCitation":"(Lamme, 2018)","noteIndex":0},"citationItems":[{"id":13159,"uris":["http://zotero.org/users/1940082/items/XG3IPEW4"],
"itemData":{
"id":13159,
"type":"article-journal","abstract":"...",
"citation-key":"lamme2018",
"container-title":"Philosophical Transactions of the Royal Society B: Biological 
Sciences","DOI":"10.1098/rstb.2017.0344","issue":"1755","page":"20170344","PMID":"30061458","title":"Challenges for theories of consciousness: seeing or knowing, the missing ingredient and how to deal with panpsychism.","volume":"373","author":
[{"family":"Lamme","given":"VAF"}],"issued":{"date-parts":
[["2018"]]}}}],"schema":"https://github.com/citation-style-language/schema/raw/master/csl-citation.json"}}

2. Pandoc parse this reference and outputs the ID, not the cite key and so any conversion via citeproc will fail:

pandoc -f docx+citations -t native test.docx
[ Para
    [ Str "This"
    , Space
    , Str "is"
    , Space
    , Str "a"
    , Space
    , Str "test"
    , Space
    , Cite
        [ Citation
            { citationId = "13159"
            , citationPrefix = []
            , citationSuffix = []
            , citationMode = NormalCitation
            , citationNoteNum = 0
            , citationHash = 0
            }
        ]
        [ Str "(Lamme," , Space , Str "2018)" ]
    , Str "."
    ]
]

This is a relevant issue: #10550 and I created a lua filter to handle this case, this filter assumes the YAML references are present in the metadata:

https://github.com/iandol/dotpandoc/blob/master/filters/citation-key.lua

@jgm -- Pandoc could check for ID and cite-key and prefer the cite-key natively I assume and it would make this workflow simpler?

My test docx:

test.docx

The Zotero ref in that docx exported as CSL JSON:

[
	{
		"id": "lamme2018",
		"type": "article-journal",
		"abstract": "Significant progress…",
		"citation-key": "lamme2018",
		"container-title": "Philosophical Transactions of the Royal Society B: Biological Sciences",
		"DOI": "10.1098/rstb.2017.0344",
		"issue": "1755",
		"page": "20170344",
		"PMID": "30061458",
		"title": "Challenges for theories of consciousness: seeing or knowing, the missing ingredient and how to deal with panpsychism.",
		"volume": "373",
		"author": [
			{
				"family": "Lamme",
				"given": "VAF"
			}
		],
		"issued": {
			"date-parts": [
				[
					"2018"
				]
			]
		}
	}
]

[iandol]

ACTUALLY, I think I am partially wrong here. While it is true that the cite-key is not used, at least if you use citeproc this doesn't matter as the embedded citation data is used:

> pandoc -f docx+citations -s --citeproc --csl nature.csl -t plain test.docx


This is a test¹.

1. Lamme, V. Challenges for theories of consciousness: seeing or
knowing, the missing ingredient and how to deal with panpsychism.
Philosophical Transaction

THEREFORE you can use citeproc without even needing a bibliography file and get out a formatted citation in LaTeX (see how CSL is changing the style for us):

> pandoc -f docx+citations --citeproc --csl nature.csl -t latex test.docx
This is a test\textsuperscript{1}.

\protect\phantomsection\label{refs}
\begin{CSLReferences}{0}{0}
\bibitem[\citeproctext]{ref-13159}
\CSLLeftMargin{1. }%
\CSLRightInline{Lamme, V.
\href{https://doi.org/10.1098/rstb.2017.0344}{Challenges for theories of
consciousness: seeing or knowing, the missing ingredient and how to deal
with panpsychism.} \emph{Philosophical Transactions of the Royal Society
B: Biological Sciences} \textbf{373}, 20170344 (2018).}

\end{CSLReferences}
                                                                                                                                                                         

> pandoc -f docx+citations --citeproc --csl apa.csl -t latex test.docx
This is a test (Lamme, 2018).

\protect\phantomsection\label{refs}
\begin{CSLReferences}{1}{0}
\bibitem[\citeproctext]{ref-13159}
Lamme, V. (2018). Challenges for theories of consciousness: seeing or
knowing, the missing ingredient and how to deal with panpsychism.
\emph{Philosophical Transactions of the Royal Society B: Biological
Sciences}, \emph{373}(1755), 20170344.
\url{https://doi.org/10.1098/rstb.2017.0344}

\end{CSLReferences}

---

If you don't use citeproc then the problem arises:

pandoc -f docx+citations --biblatex -t latex test.docx
This is a test \autocite{13159}.

Now the cite-key is the ID and it will not link the the BibTex exported by Zotero...

---

TLDR: add --citeproc and you should get formatted refs in your conversion.

[tommyvct]

Thanks for the insight, and adding --citeproc did make some difference.

But it seems that overleaf does not like this LaTeX document at all:

I got the PDF compiled anyway and it looks alright. However, if one look closely why there are so many errors, it seems that Overleaf can't find CSLReferences package, and it also seems that solving this error is also a non-trivial task.

\begin{CSLReferences}{0}{0}
\bibitem[\citeproctext]{ref-2486}
\CSLLeftMargin{{[}1{]} }%
\CSLRightInline{A. Lascarides and N. Asher, {``Segmented Discourse
Representation Theory: Dynamic Semantics With Discourse Structure,''} in
\emph{Computing Meaning}, H. Bunt and R. Muskens, Eds., Dordrecht:
Springer Netherlands, 2007, pp. 87--124. doi:
\href{https://doi.org/10.1007/978-1-4020-5958-2_5}{10.1007/978-1-4020-5958-2\_5}.}

\bibitem[\citeproctext]{ref-2495}
\CSLLeftMargin{{[}2{]} }%
\CSLRightInline{C. Li, {``Facing Data Scarcity in Dialogues for
Discourse Structure Discovery and Prediction,''} Université de Lorraine,
2023.}

The citations aren't in BibTex's \cite{} commands hence there is no internal logical link between the actual citation in text and the bibliography section.

 Segmented Discourse Representation Theory (SDRT),
developed by Asher and Lascarides (2003){[}1{]}, addresses these
limitations by adopting a graph-based representation that can
accommodate the complex structural patterns inherent in conversational
discourse .

Maybe what I want is not what this project designed to do, but ultimately I wanted DOCX in, .tex and bibtex(either embedded in .tex or standalone) out, with proper citations converted to \cite{} in the content.

[tommyvct]

PS. I noticed that you added --csl param with pandoc. I downloaded ieee.csl and I attempted conversion again with this:

pandoc -f docx+citations -t latex -i Thesis\ Proposal\ Real.docx -o proposal2.tex --extract-media=. --citeproc --csl ieee.csl

[jgm]

Overleaf can't find CSLReferences package

The CSLReferences environment is defined in the default pandoc template for LaTeX.
The relevant definition should be in the .tex file produced by pandoc (assuming you used -s and didn't use a custom template).

[jgm]

Also, if you want links from the citations, set link-citations to true in variable or metadata.
e.g. -Vlink-citations.

[iandol]

OK, I tried with your docx -- you need to specify --standalone and also use xelatex/lualatex and it does compile well (-s is same as --standalone):

> pandoc -f docx+citations --extract-media=./media --citeproc --pdf-engine=xelatex --csl apa.csl -s -t latex -o out.pdf thesis.docx

I used apa.csl just to test that citations are formatting as expected and they are:

---

However if what you want is "raw" LaTeX citations like \cite{li2023} with cite keys from BBT it will not work with your document as there are no citekeys in the Zotero metadata

[iandol]

Sorry I just realized that I missed your reply above on where I uploaded the DOCX. Now the question is, can Pandoc export the TeX with the \cite{randomID} and also export the references in BibTeX that uses the said random ID as the key?

Kind of. If you convert your docx to markdown you get something like this:

> pandoc -s --verbose -f docx+citations --extract-media=./ -t markdown thesis2.docx
[INFO] Extracting media/image1.png...

---
references:
- author:
  - family: Aru
    given: Jaan
  - family: Larkum
    given: Matthew E
  - family: Shine
    given: James M
  citation-key: aru2023
  container-title: Trends in Neurosciences
  DOI: 10.1016/j.tins.2023.09.009
  id: 23083
  issue: 12
  issued: 2023
  page: 1008--1017
  title: The feasibility of artificial consciousness through the lens of
    neuroscience
  type: article-journal
  volume: 46
---

# Chapter 1 Introduction

## 1.1 Background

Take this example:

![[]{#_Ref214896610 .anchor}Figure 1 A dialogue instance](./media/media/image1.png){width="2.46in" height="0.81in"}

In [Figure 1](#_Ref214896610), lorum ipsum[@23083].

Note how the in-text cite is a proper pandoc cite [@23083] and the reference itself is present in the YAML metadata with correct ID. But this YAML is not BibTeX. If you pipe to a second pandoc to generate LaTeX directly it will "lose" the references YAML. So in this case probably you need two steps, not a pipe:

STEP 1 (only extract the references):

pandoc -s --verbose -f docx+citations -t biblatex thesis2.docx > refs.bib
@article{23083,
  author = {Aru, Jaan and Larkum, Matthew E and Shine, James M},
  title = {The Feasibility of Artificial Consciousness Through the Lens
    of Neuroscience},
  journal = {Trends in Neurosciences},
  volume = {46},
  number = {12},
  pages = {1008–1017},
  date = {2023},
  doi = {10.1016/j.tins.2023.09.009}
}

We now have the references extracted to a BibLaTeX file. Now a second pass can use biblatex (though why you don't want to use citeproc I don't know?)

> pandoc -s --verbose -f docx+citations --extract-media=./ --biblatex --bibliography=refs.bib -t
 latex -o out.tex thesis2.docx

You now have refs.bib and out.tex that can be compiled:

> latexmk -logfilewarnings -interaction=nonstopmode -f -pv -time -xelatex ./out.tex

[tommyvct]

(though why you don't want to use citeproc I don't know?)

Because --citeproc is making no practical difference for me.

With --citeproc: pandoc -f docx+citations -i Thesis\ Proposal\ Real.docx --extract-media=./media --citeproc -Vlink-citations --standalone -t latex -o proposal.tex gives:

The study of discourse thus reveals the underlying coherence of
a text, explaining not merely what is said in each sentence, but how
those sentences work together to convey a unified communicative
intent(C. Li 2023).

Without --citeproc, pandoc -f docx+citations -i Thesis\ Proposal\ Real.docx --extract-media=./media -Vlink-citations --standalone -t latex -o proposal.tex gives:

The
study of discourse thus reveals the underlying coherence of a text,
explaining not merely what is said in each sentence, but how those
sentences work together to convey a unified communicative intent{[}2{]}.

With the latter form, I can at least use prompt engineering to replace the malformed citations, and that's unfortunately what I did ultimately.

[tommyvct]

And the latter number is exactly what shown and printed from the DOCX in question. So, with a prompt like this:

Read the #file:ref.pdf, notice that there is a number in [] in each citation. Read /Users/tommyvct/Desktop/tex/references.bib  , it's a bibtex file. Match the number with the bibtex id tag. The bibtex file is large so please, search in the file using the info from /Users/tommyvct/Desktop/tex/ref.pdf then read around what you find. 

where ref.pdf is the References section from the original DOCX, and references.bib is everything in my Zotero library exported.

The result of this query should give out a table that match the number in DOCX with the correct BibTex id. Then I can ask it further to replace the {[}2{]} to a real \cite{}:

This is the TeX document, Replace everything like `{[}16{]}` to proper `\cite{bibtex_id}` where this 16 is the citation number, and you can find the corresponding bibtex_id in mapping.md . For example. `{[}16{]}` should become `\cite{iruskieta_rst_nodate}`

where mapping.md contains the table generated from the first query.

[iandol]

Because --citeproc is making no practical difference for me.

With --citeproc ...

You are getting a properly formatted final document across multiple pandoc outputs, i.e. a document you can submit. There are some cases where BibLaTeX may offer some edge case features that citeproc doesn't, but the point is to get a formatted bibliography surely?

A single command gets your IEEE refs correctly in the PDF, why are you worried that the LaTeX source is formatted, that is by design:

I can at least use prompt engineering to replace the malformed citations, and that's unfortunately what I did ultimately.

If you have a reason why citeproc formatting is not suitable, you can still use the TWO STEP method to get to a LaTeX+BibTeX formatted PDF, so again I don't quite get why you don't do that. STEP 1: pandoc makes a .bib from the docx-embedded refs, using ID. STEP 2: use that bib for final formatting. I can compile your docx letting BibLaTeX do the formatting rather than citeproc without having to run anything other than pandoc, with your LaTeX source using \autocite{} commands:

... as later mentioned in this proposal and identified by
Li\autocite{2495}, Will the new Discord dataset alleviate this issue?
\textbf{Third}, the two-step pipeline (structure extraction then
relation prediction) proposed by Li in 2023\autocite{2495} may not be
optimal; ...

You have two methods that work without any external tools...

[tommyvct]

I realized that there is a huge difference between what Pandoc is designed to do and what I want.

You are right, it did get a job done on converting to a properly formatted final document. It does look legit. However, this document is not finalized for me because I will need to extend my thesis on top of this thesis proposal. If I want a PDF I will use Word to export PDF and not bother with this. What I actually want is a workable, minimal TeX project, that can be used in Overleaf or other TeX IDE, transferred smoothly, frictionless from DOCX. The converted TeX file contains too much other stuff that I don't care or that I am not familiar with, like the said \autocite{} and CSLReferences which I have never seen before, as I have only used LaTeX for like once or twice, and followed what exactly in the example provided by Overleaf. I know this sounds super noob and you could choose to refuse to understand this, but this is the mindset of many, many people who just need to get started from their old DOCX to LaTeX.

The other reason why I "refuse" to use this 2-step method, is I already have the exported BibTeX file exported from Zotero and I would be citing against that. even this extracted BibTeX file is a proper subset of this large exported one, they do not share the same key. This re-keying process is easier by the dumb method executed by prompt engineering.

To help you understand what I am talking about, think I want to translate a.java to a.cs. What comes out of Pandoc is equivalent to CIL byte code to me, while what I want is a logically equivalent and human workable C# source code.

I chose Word back then because I want to get things started quickly. That DOCX did serve me well for the purpose of serving a thesis proposal. However I got roasted by my supervisor for using Word, which he was kind of right on that. Word works fine if you just need to get started on writing but if you want anything fancy, like citations, it is a disaster.

And as a closing note, I strongly recommend you to do improve on the documentation, especially the demo part. Make it a scenario-based, with solutions and rendered pics, just like what we exchanged in this discussion.

Thanks for your help and dedication for helping me out, regardless the outcome.

[iandol]

I realized that there is a huge difference...

No, Pandoc is suitable for what you want, but I think I now understand your problem better.

The converted TeX file contains too much other stuff that I don't care or that I am not familiar with, like the said \autocite{} and CSLReferences which I have never seen before

1. Do not worry about \autocite{} it is standard "smart" citation and commonly used and supported for ages: https://bastian.rieck.me/blog/2016/biblatex_superscript_citations/#:~:text=The%20%5Cautocite%20command%20is%20my%20best%20friend,on%20your%20language%20settings.%20This%20is%20great. -- if you still do NOT want to use the latest LaTeX recommendation then specify the older --natbib rather than --biblatex in the pandoc command and you get older \cite{} markup...
2. CSLReferences is just visual styling when using --citeproc and not anything too exotic, I relaise if you are a TeX newbie everything is exotic 🤣

BUT I now understand that you do not want to write in Word but want a one-off conversion. In this case I agree that --citeproc is NOT the correct solution!

The TWO STEP solution is STILL the correct one that makes \[auto]cite{#} markup, but hang on to read below why (and why you can skip STEP 1):

To help you understand what I am talking about, think I want to translate a.java to a.cs. What comes out of Pandoc is equivalent to CIL byte code to me, while what I want is a logically equivalent and human workable C# source code.

This analogy is not correct. The reason is you did not use BetterBibTeX when you wrote in Word. There is no cite key there at all. It is like you expect if a.java does not declare int FOO; then you somehow expect a.cs to have int FOO declared. If you had used BBT in Zotero then my Lua filter would have worked for you. This is a weakness of Zotero, it should IMO support cite-keys as standard and/or better handle its unique numeric IDs like other reference managers do...

---

BUUUUUUT: this is still a fairly easy fix: Zotero has two APIs (web and RPC) so you should be able to write a small script to parse the LaTeX for \cite{#} get the # unique ID then retrieve the associated cite-key (now that BBT is installed) via the API, then replace # with the key. You then don't need my STEP 1 any more, the output of STEP 2 is what your script alters to give you final LaTeX markup using standard citation tools and cite-keys that are in your zotero database you can import to Overleaf etc.

[iandol]

this is still a fairly easy fix

I got Gemini to write a small script to search using the API:

https://gist.github.com/iandol/fb6ad60030e3ffcb07bdee4a5aa3557b

For example to get a cite-key from a numeric id (Zotero must be running, needs jq and curl tools installed):

> zoteroSearch '[["itemID", "is", "12187"]]' | jq '.result[0].citekey'

storm2024

You can make a second script to parse TeX, then use this to search and then replace each # with the citekey...

[tommyvct]

2. I relaise if you are a TeX newbie everything is exotic 🤣

Exactly. Thanks for bring that up, I thought that was related to the CSL thing.

Zotero has two APIs (web and RPC)

I was using this for the Overleaf. It came with small app that uses the API from Zotero cloud that gets everything into a giant bibtex file, so it have nothing to do with BetterBibTeX. I of course didn't dig deeper, but the bibtex file I get out of this apparently have a key:

@misc{chan_chatgpt_2024,
	title = {{ChatGPT} {Evaluation} on {Sentence} {Level} {Relations}: {A} {Focus} on {Temporal}, {Causal}, and {Discourse} {Relations}},
	shorttitle = {{ChatGPT} {Evaluation} on {Sentence} {Level} {Relations}},
	url = {http://arxiv.org/abs/2304.14827},
	doi = {10.48550/arXiv.2304.14827},
	urldate = {2026-04-21},
	publisher = {arXiv},
	author = {Chan, Chunkit and Cheng, Jiayang and Wang, Weiqi and Jiang, Yuxin and Fang, Tianqing and Liu, Xin and Song, Yangqiu},
	month = jan,
	year = {2024},
	note = {arXiv:2304.14827 [cs]},
	keywords = {Computer Science - Computation and Language},
}

@inproceedings{yu2022speaker,
	title = {Speaker-aware discourse parsing on multi-party dialogues},
	booktitle = {Proceedings of the 29th international conference on computational linguistics},
	author = {Yu, Nan and Fu, Guohong and Zhang, Min},
	year = {2022},
	pages = {5372--5382},
}

@inproceedings{pisarevskaya2017towards,
	title = {Towards building a discourseannotated corpus of russian},
	booktitle = {Komp'juternaja lingvistika i intellektual'nye tehnologii},
	author = {Pisarevskaya, Dina and Ananyeva, Margarita and Kobozeva, Maria and Nasedkin, Alexander and Nikiforova, Sofia and Pavlova, Irina and Shelepov, Alexey},
	year = {2017},
	pages = {201--212},
}

and this is what I need to match.